Published: July 11, 2026 · Subject: Oracle Corporation (NYSE: ORCL) and the debt-funded AI infrastructure buildout

The divergence

Two markets are currently pricing the same company, and they disagree.

Oracle’s five-year credit default swaps have moved from roughly 108 basis points in November 2025 to a record near 198 basis points by April 2026, more than doubling in five months. A spread in that upper range implies, assuming a standard 40% recovery rate, a cumulative default probability in the low-to-mid teens over five years. This repricing began the moment Oracle signed its $300 billion compute agreement with OpenAI in September 2025, and it has not moved in only one direction. Spreads briefly compressed in early February 2026 after Oracle priced a $25 billion financing package that reassured bond investors, only to blow back out to the April record within weeks. That reversal is itself informative: the relief rally lasted exactly as long as it took the market to notice that a financing plan solves near-term liquidity timing, not the underlying counterparty concentration sitting inside the backlog. The wider structural move, not the interim relief, is what has persisted.

A third reference point sat between those two until days before publication. Moody’s still rates Oracle Baa2 with a negative outlook. S&P, which had held BBB with a negative outlook through the entire CDS repricing, cut the rating to BBB- in the week of this note’s publication, the last rung of investment grade, citing rising business risk and weaker cash flow. Alongside the downgrade, S&P now projects a fiscal 2027 free operating cash flow deficit of roughly $42 billion, nearly double its prior estimate. The sequence is the point: the CDS market repriced first, the rating agency followed nine months later, and the equity consensus has done neither. The slow end of the adjustment has started to move. Equity analysts, it turns out, are slower than either.

Equity analysts, meanwhile, maintain a Strong Buy consensus on the same issuer, with a mean price target in the $252–264 range against a share price near $140, an implied upside of roughly 75%. That figure deserves no analytical weight. The bulk of those targets were set between March and June 2026, before the July correction in AI infrastructure names, and the post-correction revisions have only started. A consensus built on a market regime that no longer exists is not a bullish signal. It is a lagging indicator wearing a bullish costume.

The interesting question is not which market is right about Oracle. It is what the credit market’s early move tells us about the entire debt-funded AI buildout, three weeks before Microsoft and Meta update their capital expenditure plans. At issuers under credit stress, the CDS market has repeatedly moved before the equity market. This note takes that pattern seriously.

The mechanics: who actually pays for the buildout

The AI infrastructure trade has quietly changed its funding model. Through 2024, hyperscaler capex was financed almost entirely from internal cash generation. That era is ending. Morgan Stanley expects AI-related debt issuance to roughly double toward $570 billion in 2026 and hyperscaler spending to cross $1 trillion by 2027.

Oracle sits at the sharp end of this shift, not because it runs the weakest business, but because it is the marginal borrower of the trade. The fiscal 2026 numbers, published June 10, describe the position plainly. Capital expenditures rose 162% to $55.7 billion. Free cash flow swung to negative $23.7 billion, from a deficit of just $394 million a year earlier. Management guided fiscal 2027 net capex to roughly $70 billion and announced a further $40 billion raise in debt and equity, on top of the $43 billion of debt and $5 billion of equity raised during fiscal 2026. Total borrowings now stand near $167 billion.

Unlike Amazon, Alphabet, or Microsoft, Oracle cannot fund this buildout from operating cash flow. Every incremental gigawatt of capacity is financed on the balance sheet, in advance, against contracts whose revenue arrives years later. That sequencing is the entire story: Oracle carries the credit risk of its own order book. Which is precisely why the bond market, whose job is pricing balance-sheet risk, moved before the equity market, whose reflex is pricing growth.

The concentration problem

Oracle ended fiscal 2026 with $638 billion in remaining performance obligations, up 363% year over year. The headline is extraordinary. The composition is the risk.

According to Bank of America, more than half of that RPO is attributable to a single customer: OpenAI. The September 2025 compute agreement alone accounts for a reported $300 billion of it; BofA’s higher figure suggests additional commitments beyond that contract. Consider what stands behind that exposure. OpenAI ended 2025 at an annualized revenue run rate of roughly $20 billion, projects revenue of $280 billion by 2030, and expects to burn through more than $650 billion in cash along the way. It has made compute commitments on a comparable scale to several other cloud providers simultaneously, and its IPO, once expected earlier, is now reported by The New York Times to be under consideration for 2027. These are reported figures and reported plans, not audited outcomes; the point is not that they are wrong, but that a receivable of this size currently rests on them.

Oracle itself has told investors as much. The fiscal 2026 annual report, filed June 10, contains unusually specific risk language for a company of this size: it warns that some customers may be highly leveraged and could fail to pay or perform, that certain OCI offerings concentrate revenue among a small number of large customers in ways that amplify this exposure, and that the company may be overestimating demand and might be unable to re-lease or repurpose AI data center capacity if a customer falls away. When an issuer documents the counterparty risk on its own backlog with this level of precision, the credit market’s near-record spread stops looking like pessimism and starts looking like reading comprehension.

The equity framing treats $638 billion of RPO as deferred revenue. The credit framing treats it as a concentrated, unsecured exposure to a pre-profit counterparty, funded in advance with borrowed money. Both descriptions are technically accurate. Only one of them prices the difference between a contract and a collection.

Management’s strongest rebuttal deserves the floor. Oracle discloses that $75 billion of the large AI contracts is already covered by customer prepayments or customer-supplied hardware, which materially reduces the capital it must raise against that portion of the backlog. The mitigant is real, and it is bounded: it covers the hardware, not the buildout. The remaining obligation still requires Oracle to finance construction years ahead of collection, and $75 billion against $638 billion leaves the proportion of the order book resting on future counterparty fundraising largely intact.

The rhyme: vendor financing, 1999–2001

The structure has a precedent. In the late-1990s telecom buildout, equipment vendors extended credit to their own customers to finance purchases of their own equipment. Backlogs hit records at the exact peak of the cycle. When the customers, many of them pre-profit carriers built on capital-market access rather than cash flow, lost that access, the backlogs did not convert. They evaporated, and the vendors were left holding both the receivables and the debt raised to serve them.

The rhyme with today is structural, not cosmetic: a supplier’s balance sheet standing behind its own demand, counterparty concentration among cash-burning customers, and a market narrative that reads backlog size as backlog quality. The differences are equally structural and should be stated plainly. End demand for AI compute is real and monetizing today, not speculative. The largest builders, Microsoft, Alphabet, Amazon, and Meta, fund from operating cash flow at a scale the 1999 carriers never approached. And Oracle’s counterparty, whatever its burn rate, has revenue growing at a pace no 1990s carrier ever achieved. The analogy is a risk framework for one funding structure at the margin of the trade, not a forecast of sector collapse. Used that way, it earns its place. Stretched further, it would be narrative outrunning numbers.

What to watch: the July test

This note publishes three weeks before the thesis gets its first live test. The signals worth tracking, in order of information value:

The hyperscaler capex prints, July 29. Microsoft reports its June quarter on July 29, with Meta scheduled for the same day, and both will update capital expenditure guidance. The number matters less than the funding language. Capex raised and funded from operating cash flow extends the buildout on the old model. Capex raised alongside new debt issuance, or softened with talk of “discipline” and “phasing,” confirms that the funding constraint has become the binding one. Watch the balance sheet commentary, not the headline figure.

The credit tape. Oracle’s five-year CDS and cash bond spreads are now the cleanest real-time gauge of how professional risk-takers price the AI buildout’s weakest funding link. Spread stability through the earnings window would suggest the July equity correction overshot. Further widening, particularly if it spreads to CoreWeave-type pure plays or shows up in the terms of new AI-linked issuance, would suggest it has not finished.

Conversion, not accumulation. From Oracle’s September quarter onward, the metric that matters is the rate at which RPO converts to billed revenue, and the size of customer prepayments against future capacity. A backlog that converts is an asset. A backlog that compounds faster than it converts is a claim on someone else’s future fundraising. Disclosure changes around customer concentration would be the earliest warning of all: companies expand risk language before they restate numbers.

The OpenAI funding calendar. Every reported capital raise, IPO timing update, or new compute commitment shifts the probability distribution under more than half of Oracle’s backlog. This is the uncomfortable core of the trade: the marginal variable behind a $638 billion order book belongs to a private company that publishes no financial statements.

Conclusion

The equity market spent the first half of 2026 debating whether AI infrastructure demand is real. That was the wrong debate; the demand is visible and monetizing. The credit market has moved on to the right one: whether the weakest-funded commitments behind that demand are bankable. Having roughly doubled since November and set a record in April, Oracle’s CDS says the bond market has already answered provisionally. The late-July earnings window will show whether the equity market catches up, or whether the two continue to price different companies wearing the same ticker.

The question for August is not whether the backlog is big. It is whether it is bankable.

This framework now has fixed appointments with reality. Spinel will revisit it after the July 29 hyperscaler prints, and again when Oracle’s September quarter shows the first clean conversion data. Whichever way the signals resolve, the follow-up will say so plainly, including if they resolve against the concerns laid out here. A framework that cannot be revised by evidence is a narrative, and this publication does not sell narratives.

This note is independent research published for informational purposes only. It is not investment advice, and nothing in it constitutes a recommendation to buy or sell any security. Figures are sourced from company filings and press reports as dated in the text.

Executive Summary

ServiceNow has lost more than half its market value over twelve months. In the same period, it reported the highest operating margin quarter in its history, raised full-year guidance by $205 million at the midpoint, grew its AI revenue target from $1 billion to $1.5 billion, and saw customers spending over $1 million annually on its AI products grow 130% year-over-year. The stock is now priced at 23x forward earnings with a PEG ratio below 1.0 on a business compounding revenue at 21% with 44% free cash flow margins.

The market’s case for the derating rests on three items from the Q1 2026 earnings call: deal timing delays in the Middle East (~75 basis points of revenue headwind, characterized by management as deferred not lost, with several already closed in Q2), margin pressure from the Armis acquisition closed April 20 (125 basis points on operating margin for Q2, absorbed over the course of 2026), and residual federal exposure. None of the three is structural.

At $93.46, ServiceNow is priced as though the AI infrastructure buildout it sits at the center of has stopped, and as though the enterprise workflow platform with a 54% Rule of 50 score deserves the same multiple as a business growing at half the rate. In my view, it does not. I am long ServiceNow, position initiated at publication of this note. The next datable test is Q2 2026 earnings on July 22.

The Setup

ServiceNow traded at $211.48 twelve months ago. It trades at $93.46 today. In the same interval, the company grew subscription revenue by 22%, posted the highest operating margin quarter in its history, raised full-year guidance three times, and built one of the most commercially validated AI product lines in enterprise software. The stock is now within thirteen dollars of its 52-week low of $81.24.

The most acute moment in the derating was April 23, 2026, the day after Q1 2026 earnings. ServiceNow reported $3.77 billion in revenue, growing 22% year-over-year. Current remaining performance obligations came in at $12.64 billion, 100 basis points above the company’s own guidance. Non-GAAP operating margin printed at 32%, the highest single-quarter figure in the company’s history, and free cash flow margin reached 44%. Full-year subscription revenue guidance was raised by $205 million at the midpoint. The stock fell 17% the following day, the worst single-session decline in ServiceNow’s fourteen-year history as a public company.

The Q1 reaction did not happen in isolation. ServiceNow had already lost roughly 30% of its market value in the first four months of 2026 before that print. The Q1 drop was the most visible expression of a derating that had been underway across the entire period, driven by a combination of sector-wide multiple compression in enterprise software, concerns about margin trajectory following the $7.75 billion Armis acquisition, and what the market read, incorrectly in my view, as a guidance structure that relied more on currency tailwinds and inorganic contribution than on organic acceleration.

The result is a business that has compounded revenue above 20% for twelve consecutive quarters, generates $4.6 billion in annual free cash flow, and is guiding to $30 billion in subscription revenue by 2030, now trading at a forward earnings multiple that has compressed 78% from its three-year average.

What the Business Actually Looks Like

ServiceNow operates a single platform (the Now Platform) that functions as the workflow orchestration layer for large enterprises. It does not sell point solutions. It sells the system through which IT operations, employee experience, customer service, security, and increasingly AI agents are coordinated, monitored, and resolved. The platform’s structural advantage is switching cost: once ServiceNow is the system of record for enterprise workflows, replacing it requires migrating years of accumulated process logic, integrations, and institutional knowledge. Renewal rates reflect this. The business does not churn.

On the core workflow business: current remaining performance obligations ended Q1 2026 at $12.64 billion, up 22.5% in constant currency, 100 basis points above the company’s own guidance. Total RPO reached $27.7 billion, up 25% year-over-year. These are contracted future revenues, already signed, already in the backlog. The subscription revenue guidance for 2026 ($15.735 to $15.775 billion) is not a forecast built on assumptions about new deal flow. It is largely already contracted.

On the AI layer: Now Assist, ServiceNow’s AI product line, entered Q1 2026 at $750 million in annual contract value, having crossed $600 million in 2025 and more than doubled year-over-year. Management raised the full-year AI ACV target from $1 billion to $1.5 billion during the Q1 call. Customers spending over $1 million annually on Now Assist products grew 130% year-over-year. Sixteen transactions exceeded $5 million in net new ACV in Q1, growing 80% year-over-year, indicating that AI adoption is expanding into larger enterprise commitments, not just broad adoption at the margin. The average ACV of renewing Now Assist customers expanded by 3x in 2025.

The monetization model is straightforward. ServiceNow bundles AI capabilities inside existing workflow subscriptions rather than selling AI tokens separately. Customers adopting AI-infused bundles see a 20–30% price uplift at renewal. This is not speculative AI revenue: it is incremental pricing on the existing installed base, flowing through the same renewal cycle that already drives the RPO trajectory above.

At Knowledge 2026 in May, ServiceNow announced a major expansion of its Autonomous Workforce, extending AI specialists capable of completing end-to-end processes autonomously across IT, CRM, HR, finance, legal, and security. The commercial results already in the field are concrete: 91% of cases resolved without human reassignment across the installed base, Docusign targeting 90% autonomous resolution of IT tickets, and the city of Raleigh reporting a 98% deflection rate on employee requests. These are not pilot metrics. They are production results on a platform already running over 100 billion workflows annually. The Autonomous Workforce expansion is the mechanism by which the 20–30% renewal price uplift compounds into operating leverage over time.

The Armis acquisition, closed April 20, adds a third layer. At $7.75 billion in cash, it brought in a cybersecurity asset exposure management platform with over $300 million in ARR and real-time visibility across physical, virtual, and cloud environments. The strategic logic is that as AI adoption expands enterprise attack surfaces, real-time asset visibility and automated remediation become infrastructure-layer requirements rather than optional security tools. ServiceNow is positioning the combined platform (workflow orchestration, AI governance via AI Control Tower, and security exposure management) as the operating system for enterprise AI deployment. The total addressable market for security and risk solutions more than triples with Armis included.

The IBM partnership, expanded on June 11, adds a fourth dimension. ServiceNow and IBM deepened their integration across IBM watsonx, Red Hat, and IBM’s automation stack, targeting the two most persistent blockers in enterprise AI deployment: AI-ready data and legacy application modernization. Early deployments show 40% faster incident resolution. IBM’s enterprise customer base in financial services, healthcare, and government represents a distribution channel into accounts where ServiceNow’s direct sales motion is slower. A set of pre-built joint blueprints for IT operations, HR service delivery, and customer service rolls out through H2 2026.

The margin structure, even absorbing Armis dilution, remains exceptional. Non-GAAP operating margin guidance for the full year sits at 31.5%, down from 32% in Q1 but still among the highest in large-cap enterprise software. Free cash flow margin is expected at approximately 32–33% for 2026 after the Armis headwind. The company repurchased $2.2 billion of stock in Q1 alone, with $4.2 billion remaining under the existing authorization, not the capital allocation behavior of a management team that believes the thesis is impaired.

Why the Market Sold It Anyway

The Q1 2026 sell-off was not irrational in its inputs. It was irrational in its conclusions. Three specific items drove the reaction, and none of them describes a structural change in the business.

Middle East deal deferrals. Several large on-premise deals in the Middle East did not close in Q1, creating approximately 75 basis points of headwind to subscription revenue growth. Management was explicit on the earnings call: the deals were deferred by one to two quarters due to customer decision timelines disrupted by the regional conflict, not lost. By the time of the Q1 call, several had already closed in Q2. A 75 basis point headwind to one quarter of revenue growth, on a business with $27.7 billion in total RPO and 22.5% constant-currency cRPO growth, does not explain a 17% single-session decline in a $100 billion market capitalization company. The market treated a timing item as a demand signal. It was not one.

Armis margin dilution. The $7.75 billion acquisition of Armis closed on April 20, four days before the earnings call. The deal introduces 125 basis points of operating margin headwind in Q2 and 75 basis points for the full year 2026, and approximately 200 basis points of free cash flow margin pressure for 2026. These are real and visible numbers, but they are one-time integration costs on an asset that triples ServiceNow’s total addressable market in security. The correct frame is not margin compression: it is upfront investment with a defined horizon. By 2027, Armis contributes approximately 125 basis points to subscription revenue growth while the integration dilution is fully absorbed. The market read the cost without reading the return.

Federal exposure and residual headwinds. Federal budget uncertainty and government spending caution contributed additional deal timing noise in the quarter, compounding the geopolitical headwinds already present in the Middle East. This was the smallest of the three items and the least discussed, but it contributed to a narrative of execution uncertainty at a moment when the stock was already under pressure from the first two items. The effect is mechanical and non-recurring.

Taken individually, each drag is defensible as a reason for caution at the margin. Taken together, they do not explain a 50% drawdown from peak on a business that beat every financial metric in the quarter, raised full-year guidance, and posted its best-ever operating margin print. The market’s conclusion, that ServiceNow’s earnings power is structurally impaired, is not supported by the Q1 data. The market’s reaction, a 17% single-session decline followed by a sustained derating to 23x forward earnings, reflects a sentiment event, not a fundamental one. The distinction is the thesis.

The Variant Perception

The consensus position on ServiceNow is not bearish. Of 54 analysts covering the stock, 43 rate it a Buy, with an average 12-month price target of approximately $142, representing 53% upside from the current price. The issue is not that the sell-side disagrees with the thesis. The issue is that the sell-side has been wrong for twelve months running, and the gap between the average price target and the actual stock price has widened rather than closed as the business continued to execute. The market is not debating whether ServiceNow is a good business. It is debating whether the current multiple is defensible, and it has been answering that question by compressing the multiple further each time the stock rallies.

The variant perception here is not about the business. It is about the valuation floor.

ServiceNow trades at 23.6x forward earnings, 21x price-to-free cash flow on trailing twelve months, and a PEG ratio of 0.95, below 1.0 for the first time in the company’s publicly traded history. The free cash flow yield has reached 5.05%, against a 15-year historical average of 2.06%. The current forward P/E represents a 78% discount to the three-year average of 104x and an 81% discount to the five-year average of 243x. These are not minor compressions. They describe a business whose valuation has been reset from growth software to something approaching mature enterprise software, without a corresponding reset in the growth rate.

The peer comparison makes the dislocation precise. Salesforce trades at 13.9x forward earnings and grows subscription revenue at approximately 9.6% annually. ServiceNow grows at 21%, more than double the rate, and trades at 23.6x forward earnings, a 70% premium for 2.2x the growth. On a pure growth-adjusted basis, ServiceNow should command a meaningfully larger premium. The Rule of 50 metric, which combines revenue growth and free cash flow margin as a composite measure of SaaS financial health, produces a score of 54% for ServiceNow, the highest in its peer group, ahead of Workday at 42%, Salesforce at 42%, Microsoft at 42%, Oracle at 24%, and SAP at 22%. The company with the strongest composite operating performance in large-cap enterprise software is trading at a multiple that prices it as one of the weakest.

The argument against re-rating is that multiple compression is structural: that AI competition from Salesforce Agentforce, SAP Joule, and Microsoft Copilot will erode ServiceNow’s pricing power and growth trajectory before the market is willing to re-rate the stock. This is the bear case on the multiple, and it deserves to be taken seriously. What it does not explain is the current evidence: Now Assist customers spending over $1 million growing 130% year-over-year in the quarter where the stock lost 17%. Competitive displacement and accelerating AI monetization do not coexist. The business that is being disrupted does not raise its AI revenue target by 50% in a single earnings call.

The more credible read is that the current multiple reflects three things: sector-wide de-rating of high-growth software that began in late 2025 and has not fully reversed; the overhang from the Armis acquisition, which introduced real but temporary margin pressure at a moment of maximum market sensitivity; and a consensus that has been systematically revising price targets downward for three months without an anchor to stop the revision cycle. When the Q2 print resolves the Middle East deal question and demonstrates that Armis integration is proceeding within the guided parameters, the revision cycle has no remaining justification. The multiple does not need to return to 104x to generate a significant return from here. It needs to return to 30x.

Bull, Base, and Bear

The scenarios below are built primarily on a price-to-free cash flow framework, using two forward periods: FY2026 and FY2027. FCF is the cleaner valuation anchor for ServiceNow given the gap between GAAP and non-GAAP earnings from acquisition accounting and stock-based compensation. Non-GAAP EPS figures are used as a cross-check. FY2026E FCF per share is estimated at approximately $5.00, reflecting the company’s guided 32–33% annual FCF margin on $15.76 billion in midpoint subscription revenue after the Armis 200 basis point headwind. FY2027E FCF per share is estimated at approximately $6.60, reflecting margin recovery to 35–36% as integration costs are absorbed on $19.16 billion in consensus revenue.

Bull case at $185–210 (12-24 months). Requires Q2 2026 to confirm that Middle East deferrals have closed as management indicated, AI and cloud growth sustaining above 30% through H2 2026, Now Assist ACV reaching the $1.5 billion target on schedule, and Armis integration proceeding within the guided parameters without incremental dilution. Under this scenario, the multiple re-rates toward 30–32x price-to-FCF on FY2027E as the sentiment overhang from Armis clears and the AI monetization trajectory is no longer in dispute. Math: $6.60 × 30x = $198. Top of range: $6.60 × 32x = $211. Non-GAAP EPS cross-check: $5.08 × 40x = $203. The range is consistent with the upper end of current sell-side targets and requires a multiple that remains well below the three-year historical average of 104x.

Base case at $130–150 (12 months). Management delivers on the upgraded 2026 guidance, Q2 earnings confirm the Middle East deals closed and Armis margin guidance holds, but AI monetization growth decelerates modestly from Q1’s pace and the market continues to debate the correct peer group. The multiple expands from current levels to 27–30x price-to-FCF on FY2026E as the transitory drag narrative is resolved but full re-rating awaits cleaner FY2027 visibility. Math: $5.00 × 28x = $140. Non-GAAP EPS cross-check: $4.17 × 33x = $138. Range $130–150. The stock currently trades at roughly 18.6x price-to-FCF on FY2026E; reaching the base case requires the market to close less than half the gap to the three-year average multiple.

Bear case at $70–82 (12 months). Three independent paths lead here. Q2 2026 disappoints on AI and cloud growth, with deceleration below 20% suggesting Q1’s trajectory was partially front-loaded by early adopter uptake rather than durable demand expansion. Armis integration produces incremental margin pressure beyond the guided 200 basis points, forcing a second downward revision to FCF guidance and reopening the acquisition premium debate. A broader enterprise software spending pause, driven by macro deterioration or a hyperscaler capex signal, reduces renewal expansion rates and compresses the cRPO growth trajectory. Under any of these paths, the multiple stays pinned at 14–16x price-to-FCF on FY2026E with no catalyst to break the revision cycle before year-end. Math: $5.00 × 14x = $70. Non-GAAP EPS cross-check: $4.17 × 18x = $75. Range $70–82.

Invalidation below $60. The thesis is structurally broken. A price below $60 implies less than 12x price-to-FCF on FY2026E estimates, a multiple that would require either a recession-driven collapse in enterprise software spending, a competitive displacement scenario where ServiceNow’s renewal rates deteriorate materially, or an Armis integration failure that forces a goodwill impairment and a fundamental reset of the capital allocation framework.

Catalysts

The primary datable event is Q2 2026 earnings, scheduled for July 22, after market close. This is the first print that can empirically resolve the three items that drove the April 23 sell-off. On the Middle East deferrals: management stated on the Q1 call that several of the delayed deals had already closed in Q2. A Q2 subscription revenue print that confirms cRPO growth sustained at or above 22% constant currency would close that debate. On Armis: Q2 guidance assumed 125 basis points of operating margin dilution. A Q2 actual margin at or above the 26.5% guided figure confirms that the integration is proceeding within the stated parameters and removes the market’s justification for pricing incremental acquisition risk. On AI momentum: the Q1 pace of 130% growth in Now Assist customers above $1 million ACV is the number to watch. Deceleration below 80% raises questions about whether Q1 reflected pull-forward; sustained growth above that threshold confirms the trajectory.

Hyperscaler Q2 earnings calls, running in the same window as ServiceNow’s print, will provide independent validation or a challenge to the enterprise AI spending environment. Any signal of sustained hyperscaler infrastructure capex above $700 billion for 2026 is incrementally positive for ServiceNow’s Armis and AI Control Tower positioning.

Now Assist reaching $1 billion in cumulative ACV during H2 2026, against the raised $1.5 billion full-year target, is the milestone that would shift the AI monetization narrative from promising trajectory to demonstrated scale. ServiceNow entered Q2 at $750 million in Now Assist ACV. Crossing $1 billion would put the full-year target within reach and represent one of the fastest enterprise software products to reach that milestone.

The IBM joint blueprint rollout through H2 2026 provides a second distribution catalyst, expanding ServiceNow’s reach into IBM’s installed base of financial services, healthcare, and government accounts.

A capital markets event likely in late 2026 is the venue where the 2028 guidance corridor could be revised upward if the trajectory holds, and where the Armis TAM expansion gets quantified against a revenue model. That event is outside the 12-month base case horizon but anchors the bull case.

This note will be updated after Q2 earnings.

This is a Spinel Weekly Research Note. It is not investment advice, a recommendation to buy or sell any security, or a solicitation of any kind. The full disclaimer is published with the founding note of Spinel Capital Research.

I hold a personal position in Nokia, established before the launch of Spinel Capital Research. The position is therefore not part of the Model Portfolio. See full disclaimer below.

Executive Summary

Nokia has appreciated approximately 180% over the past twelve months as the market began pricing in the Nvidia AI-RAN partnership, the Infinera optical integration, and the Q1 2026 guidance upgrade. At $13.83, the stock sits at the lower end of a 12-month base case of $13-17, with a 12-24 month bull case of $20-24, and downside to $8-11 in the bear scenario with invalidation below $7. I am long on the 12-24 month case, driven by the combination of optical infrastructure positioning, AI-RAN commercial trajectory, and longer-term 6G optionality, while acknowledging that the thesis rests on multiple sustenance and forward execution rather than on cash flows already transformed. The next datable test is Q2 2026 earnings on July 23.

Thesis

Nokia is, in my view, structurally undervalued on a 12-24 month horizon. The combination of the Nvidia AI-RAN partnership, the optical infrastructure positioning the company built through the Infinera integration, and the embedded 6G architecture work makes Nokia one of the few Western infrastructure exposures positioned for the AI cycle. The appreciation of approximately 180% over the past twelve months is the market’s first-order response to that structural shift, not its conclusion.

The variant perception that supports the view is straightforward. The consensus has now priced Nokia as an AI-adjacent telecom equipment maker. I think the market is underweighting the AI infrastructure exposure within Nokia’s mix, even as Mobile Infrastructure remains the larger segment by revenue. The growth and incremental margin is shifting toward the AI side faster than the consensus is modeling, and that distinction is worth several turns of multiple over the next 12-24 months. The remainder of this note works through what Nokia is now, where the market has caught up and where it has not, what the thesis rests on in evidence terms, and the three scenarios.

What Nokia Actually Is Now

The 2026 reorganization concentrated Nokia into two operating segments plus a holdco for divested units. The legacy phone-company associations no longer apply to either growth engine.

Network Infrastructure generated €7.65 billion of net sales in 2025 at a 10.1% operating margin. The segment combines Optical Networks at €3.02 billion (roughly half from the Infinera assets integrated in February 2025), IP Networks at €2.59 billion, and Fixed Networks at €2.03 billion. Optical and IP together are the AI infrastructure exposure. They grew 6% at the segment level in Q1 2026, with Optical specifically up 20% on a constant currency and portfolio basis. Management raised the 2026 segment guidance after Q1 from 6-8% to 12-14%, and the Optical-plus-IP subsegment guidance from 10-12% to 18-20%. The optical product roadmap visible from the May 2026 JPMorgan TMT conference includes four new DSPs and thirteen application-optimized platforms, with sampling in H1 2027 and volume in H2 2027. A new indium phosphide manufacturing facility is ramping to 25x capacity by 2027.

Mobile Infrastructure generated €11.41 billion at a 13.4% operating margin in 2025, representing 57% of group revenue. The segment combines Radio Networks, Core Software, and Technology Standards (the patent licensing business). It grew 3% in Q1 at 48.5% gross margin and 8.9% operating margin. The AI-RAN narrative sits inside Radio Networks but does not yet appear in the P&L. Customer field trials begin in the second half of 2026, with ten operators now publicly committed to working on the Nokia-Nvidia stack. Beyond AI-RAN, Mobile Infrastructure also carries Nokia’s 6G architecture work, including the joint development with Nvidia announced in October 2025 covering radio access for the 6G transition that begins commercial cycle in 2028-2030.

Portfolio Businesses holds €845 million of non-core revenue at -10.7% operating margin, to be sold or wound down. Disposal upside, near-term drag.

The structure matters for valuation framing. Mobile Infrastructure remains the larger business by revenue, and any thesis that treats Nokia as a pure AI infrastructure exposure overreaches. The right framing is a conglomerate where the AI-exposed segment is growing materially faster than the legacy segment, with the mix shifting over time.

The Inflection

October 28, 2025 was the date. Nvidia disclosed an equity investment of $1 billion in Nokia at $6.01 per share, approximately 8% of Nvidia’s 13F holdings, paired with the AI-RAN partnership covering the Arc Aerial RAN Computer, 6G architecture development, and a T-Mobile US deployment commitment. The market response was muted, sell-side coverage took weeks to integrate the partnership into models, and the original repricing window stayed open longer than the announcement justified.

The catch-up since has been ordered. MWC 2026 in early March was where the partnership stopped being a press release and became Nokia’s declared corporate identity, with operators including BT, Elisa, NTT DOCOMO, and Vodafone publicly committing to AI-RAN integration work. Q1 2026 results in late April produced the operational evidence: revenue +4%, AI and cloud revenues +49%, €1 billion of AI and cloud orders booked, gross margin +320bps to 45.5%, and a guidance raise across Network Infrastructure subsegments. Management also revised its AI and cloud addressable market assumption from a 16% CAGR to 27% through 2028, with hyperscaler capex assumptions raised to greater than $700 billion in 2026.

Then came the sell-side cascade. Northland $13 to $20 on June 3. Bank of America €11 to €14.40 on June 8. JPMorgan $14 to $21 on June 13.

A further data point arrived on June 16: Nokia announced a $30 million expansion of its Allentown, Pennsylvania advanced test and packaging operations, part of a previously disclosed $4 billion multi-year U.S. R&D and manufacturing commitment. The expansion increases photonic chip production capacity 10x at the site, with commercial availability by end of Q3 2026, materially accelerating the optical capacity timeline relative to the broader H2 2027 volume ramp visible in the May roadmap. The announcement came with CHIPS Act backing and a state-level governmental partnership in Pennsylvania. The market reaction was paradoxical: the stock fell approximately 6% on the announcement day and the pullback has not fully reabsorbed, which I read as profit-taking on an extended chart and capex digestion rather than a fundamental rejection. The substance, a 10x capacity ramp with sovereign backing, is structurally positive for the indium phosphide and integrated optical narrative that supports the thesis.

What the Market Has Repriced, and What It Hasn’t

The repricing from approximately $5 to $13.83 over the past twelve months captures three specific things. The Nvidia $1 billion investment announced October 28, 2025, sized at approximately 8% of Nvidia’s 13F holdings. The strategic positioning declared at MWC 2026 in early March, when AI-RAN moved from press release to declared corporate identity. The Q1 2026 results in late April that confirmed the trajectory with +49% AI and cloud revenue growth, €1 billion of AI and cloud orders booked in a single quarter, and the guidance raise across Network Infrastructure subsegments.

What the repricing has not yet captured, in my reading, is three things.

First, the optical roadmap. The product launches with sampling in H1 2027 and volume in H2 2027 are not yet in sell-side models because the order book has not built far enough to validate them. The Allentown capacity ramp coming live by end of Q3 2026 is the leading indicator that the volume cycle starts sooner than the consensus assumes. Ciena trades at 32x forward earnings on 24% revenue growth guidance, and Ciena is the right comp for Nokia’s Optical subsegment in isolation. Optical is one segment within Network Infrastructure (€3.02 billion of €7.65 billion), and applying Ciena-tier multiples to that portion specifically lifts the blended group multiple, but does not justify applying Ciena multiples to the entire group.

Second, the AI-RAN commercial step. The current price reflects the partnership narrative but not commercial validation. The first operator going from trial to commercial pilot is a step-change moment that the multiple should respond to, and the trial calendar runs through H2 2026 and into 2027.

Third, the 6G optionality. Nokia is one of three vendors globally (with Ericsson and Samsung at scale, Huawei excluded from Western markets) positioned for the 6G cycle that begins to commercialize from 2028. The Nvidia partnership covers both AI-RAN today and 6G architecture for the next decade. This is longer-term optionality that current sell-side models do not yet attempt to value, and that begins to enter forecast windows toward the back end of the 12-24 month horizon rather than driving it.

The Variant Perception

The 35-analyst consensus 12-month price target sits at approximately $10. The recent bullish updates (JPMorgan $21, Northland $20, BofA €14.40) are pulling that average upward but most older PTs remain stale at $5-8. The market dispersion is real because the market has not settled on which mental model applies: legacy telecom, blended optical-plus-telecom, or AI infrastructure.

My reading is that on a 12-24 month horizon, the blended Nokia multiple deserves to sit between Ericsson (19-23x) and Ciena (32x), rather than compressing back toward Ericsson alone. The right blended range is 28-32x given Mobile Infrastructure’s continued weight in the mix; Ciena’s 32x flat applies to a pure-play optical comparable, which the Optical subsegment is on a standalone basis but the group is not. The case for the blended multiple at the upper end of that range rests on three observations.

The mix is shifting fast. AI and cloud revenue grew 49% in Q1 on a small base, but management raised the AI/cloud CAGR assumption to 27% through 2028. Sustained at that pace, the AI and cloud share of group sales moves from 8% today to roughly 25% by 2028, which is a different business than what the consensus is averaging.

The optical asset is structurally scarce. Western optical infrastructure with material exposure to hyperscaler capex includes Ciena, Lumentum, Coherent, and Nokia’s Optical subsegment. Of those, Nokia is the only one with integrated IP routing, the only one with indium phosphide vertical integration ramping to 25x by 2027, and the only one with a Nvidia equity partner. That combination is worth a premium within the optical comp set, not a discount.

The 6G window is closer than the consensus assumes, even if commercial cycles begin only in 2028-2030. The standardization work for 6G accelerates through 2026-2027. Nokia, Ericsson, and Samsung have the architecture work in flight; Nokia is the only one anchored to Nvidia’s compute platform. The 12-24 month window captures the inflection where this optionality begins to enter sell-side models as a tailwind rather than a far-future consideration.

What’s Narrative and What’s Proven

The 12-24 month bull case rests primarily on multiple sustenance at AI infrastructure peer levels and on the continued mix shift toward AI/cloud revenue, not on cash flows that have already materialized at scale. The distinction matters for sizing the thesis honestly.

What is already in the accounts: a guidance raise after Q1 2026, AI and cloud revenue growing 49% year-on-year on a base of approximately €360 million (8% of group sales), gross margin expansion of 320bps to 45.5%, operating margin expansion of 200bps to 6.2%, and €1 billion of AI and cloud orders booked in a single quarter. The strategic partnership with Nvidia is in place at $1 billion of invested capital. These are real, recurring where applicable, and visible in the P&L.

What is projection: the optical product roadmap volumes are 12-18 months out, with sampling in H1 2027 and volume in H2 2027 (though the Allentown capacity ramp by end of Q3 2026 brings part of this forward). The AI-RAN commercial cycle is gated by trials beginning in H2 2026 with no commercial revenue yet. The 6G commercial cycle is 2-4 years out. The 2028 EPS bull case requires consistent execution against a trajectory that has produced one quarter of evidence.

The thesis assumes the blended multiple holds at the upper end of the 28-32x range while earnings catch up. A setback in the trajectory hits the multiple before earnings catch up, which is why the bear case is real even where the long-term case is intact. This is a forward bet on re-rating sustenance and mix shift, not a present-tense valuation case.

Bull, Base, and Bear

Bull case at $20-24 (12-24 months). Requires AI-RAN moving from trial to commercial scale with at least two major operators (T-Mobile plus one other), optical roadmap executing on the H2 2027 volume timeline with material share gains against Ciena, hyperscaler capex sustaining at or above $700 billion annually, and Mobile Infrastructure margins holding at 12-14%. 2027-2028 EPS reaches €0.58-0.65, blended multiple sustained at 30-33x as the AI infrastructure interpretation becomes the dominant lens for valuation. Math: €0.60 × 30x = €18.00, $19.80 ADR. Top of range: €0.65 × 33x = €21.45, $23.60 ADR. This range is in line with current bullish sell-side targets (JPMorgan $21, Northland $20) with modest premium for optical roadmap optionality.

Base case at $13-17 (12 months). Management hits the upgraded 2026 guidance, AI and cloud growth runs at 30-35% (deceleration from Q1’s +49% but still strong), Q2 2026 in line, T-Mobile trials proceeding without commercial step change. 2027 EPS reaches €0.45-0.55, multiple compresses modestly to a blended range of 26-30x as the consensus continues to debate Nokia’s correct peer group. Math: €0.48 × 27x = €12.96, $14.30 ADR. Range $13-17. The stock currently trades at the lower end of this range.

Bear case at $8-11 (12 months). Three independent paths could trigger this scenario. Q2 2026 misses on AI and cloud growth, with deceleration below 25% suggesting Q1’s +49% was front-loaded by integration accounting rather than underlying demand. A major hyperscaler signals capex caution on its own Q2 call, putting the >$700 billion 2026 capex assumption at risk and compressing the entire AI infrastructure complex. The T-Mobile AI-RAN trial slips into 2027 or beyond with no commercial language attached, converting narrative into execution doubt. Under any of these paths, 2027 EPS falls toward the management midpoint of €0.35-0.40, the multiple compresses to Ericsson territory at 18-22x, and the AI infrastructure premium unwinds. The technical floor in this scenario is thin: the June 5 drop of 12.5% on a routine €500 million bond refinancing and the June 16 drop of 6% on a thesis-positive announcement both illustrate how quickly momentum names sell off. Math: €0.37 × 20x = €7.40, $8.15 ADR.

Invalidation at $7 or below. The thesis is structurally broken. The multiple has compressed beneath Ericsson, the AI infrastructure premium has unwound, and either earnings have disappointed materially or the macro context has turned against AI infrastructure capex.

Top Risks

Multiple compression on a single weak print. Nokia trades at approximately 41x 2026E forward P/E. A Q2 print that disappoints on AI and cloud growth, or a single hyperscaler signaling capex caution, can produce 20-30% multiple compression in days. The June 5 and June 16 drops illustrate how thin the technical floor is when the narrative encounters friction, even when the underlying news is structurally positive.

Execution risk on AI-RAN commercialization. The AI-RAN narrative depends on customer trials converting to commercial deployments. Trials beginning in H2 2026 must produce visible operator wins by 2027. Slippage compresses the 12-24 month case meaningfully because the bull multiple only holds if the trajectory holds.

Hyperscaler capex cycle. Nokia Optical is leveraged to AI buildouts at the hyperscalers. A cyclical pause in capex, whether driven by macro conditions or strategic recalibration, hits Nokia faster than the consensus modeling implies because Optical orders convert quickly to revenue.

Mobile Infrastructure margin pressure. The base case assumes Mobile Infrastructure margins hold at 12-14%. A faster-than-expected decline in legacy 5G procurement or pricing pressure from operators could compress margins below this level and break the funding case for the AI investments.

Catalysts

The next datable event is Q2 2026 earnings, scheduled for July 23. Consensus expects €4.82 billion in revenue, €0.05 EPS, and the AI and cloud growth rate as the follow-up to Q1’s +49%. Above 30% sustains the bull path; below 25% pushes the framework toward base or bear. The Q2 call is also where T-Mobile trial commentary and hyperscaler order book momentum will be read.

Hyperscaler Q2 capex commentary, on calls in the same window, will validate or challenge the >$700 billion 2026 capex assumption.

AI-RAN customer trial outcomes through H2 2026 are where narrative becomes commercial validation. The first operator demonstration moving from announcement to commercial pilot is the marker that resolves the most uncertainty.

A capital markets refresh likely in November 2026 is the venue where the 2028 guidance corridor could be revised upward if the trajectory holds.

This note will be updated after Q2 earnings.

This is a Spinel Weekly Research Note. It is not investment advice, a recommendation to buy or sell any security, or a solicitation of any kind. The full disclaimer is published with the founding note of Spinel Capital Research.

Mispriced companies. Datable catalysts. Asymmetric setups.

Most equity research is opinion. Some of it is data. Very little of it is discipline.

Spinel Capital Research is built around a narrow and consistent discipline: identifying public companies where a clear, datable catalyst is poised to force a re-rating that the market has not yet priced over a twelve-to-twenty-four month horizon, taking positions where the downside is bounded and the upside is asymmetric, and writing the entire thesis down, including the exit, before the position is taken.

This first letter exists to make explicit what that discipline is, how I will apply it in every published deep-dive thesis, and what you can hold it accountable to. The framework is presented in three pillars, followed by the operational test that each idea must pass before it becomes a published thesis. The first live thesis follows within the next four weeks.

What Spinel Is, and What It Is Not

Spinel Capital Research is an independent equity research letter.

The universe is global and unrestricted by market capitalization. Spinel will follow opportunity wherever it sits: Europe, North America, Asia, large cap, mid cap, occasionally small cap when liquidity permits. What it will not do is dilute its conviction by writing about names that do not clear a high bar of asymmetric risk-reward.

The letter is published in two formats. Once a month, a deep-dive note dedicated to a single thesis examined at the level of detail required to actually take a position: origin of the mispricing, catalyst, valuation framework, bear case, exit framework with explicit upside target and invalidation level. Once a week, Spinel Weekly publishes shorter work in three defined categories: Research Notes, devoted to a single company that holds analytical interest but does not clear the three-pillar bar required for a Deep Dive; Industry Notes, focused on a sector observation, a thematic lens, or a cycle worth understanding; and Portfolio Notes, dedicated to post-mortems on closed positions, six-month reviews of published theses, and the written acknowledgment of what worked and what did not.

Once a month, Spinel Weekly publishes The Rejects, a short module listing three to five names that were considered and screened out, with the reason for rejection stated in two or three lines per name. You see the work that did not make the cut, which is where most of the discipline actually lives.

The deep dive is the work. The weekly is the thinking that surrounds it.

What Spinel will not do is equally important. The deep-dive mandate covers long-only equity, single-name, catalyst-driven theses. It will not cover cryptocurrency, private equity, venture capital, or any asset class outside listed equities, and it will not recommend short positions, options strategies, or leveraged products. The Weekly format is wider in scope, but the underlying philosophy remains the same: rigorous, written analysis, no hot takes, no signal flow, no momentum chasing. Macro reflections and sector observations are welcome where they sharpen the work; predictions about indices and market timing calls are not.

The track record will be public, but it will not be my personal trading account. Every deep-dive thesis published in Spinel will be tracked in a single, transparent Model Portfolio, with entry prices set at the close on the day of publication. The Model Portfolio exists to provide a fully auditable record of the research over time: every winner, every loser, with no cherry-picking and no rewriting of history.

A word on why the Model Portfolio is not my personal book. I am subject to professional constraints that affect personal trading. Certain names covered in these pages cannot be held personally. Others can, in sizes and timings that may differ materially from the model. Conflating the two would compromise both: the research, by narrowing the universe to what I can trade, and the personal book, by exposing it unnecessarily. The Model Portfolio is the discipline of the work made visible. Where my personal positions stand at any given time is not the subject of this letter.

Three Pillars

Every investment discipline has to answer one question: where does the edge come from? Without a clear answer, what looks like skill is usually just exposure to a market regime that happened to work. The work of Spinel rests on three answers, each one a pillar without which the discipline cannot stand.

Contrarian: Mispricing from the Collapse of Attention

Markets are not always wrong. They are wrong in specific, recurring ways. The most reliable of these is the collapse of attention. When a company falls out of the narrative (when sell-side coverage thins, when retail interest moves elsewhere, when the sector becomes structurally uninteresting to allocators), the price detaches from the underlying business in a way that the price of well-followed companies rarely does. That detachment is the mispricing Spinel is built to find.

The contrarian pillar is not about being negative on what the market loves, or positive on what the market hates. It names a specific source of mispricing (the absence of competition for information), not a posture of disagreement with consensus. A position that requires the entire market to be wrong about something obvious is a bad position. A position that only requires the market to start paying attention again is a good one.

Catalyst-driven: A Datable Reason to Be Right

A cheap stock that stays cheap is not a thesis. It is a value trap. The bridge between mispricing and return is a catalyst: a specific event, change of cycle, or shift in narrative that forces the market to update its assumptions. Without one, the position has no horizon and no falsifiability. With one, it has both.

A catalyst is not merely an event. It is an event capable of changing the market's expectations. Most quarterly results, most product launches, and most management appointments are events that pass without forcing a reassessment. The distinction matters: the test of a catalyst is not whether it occurs, but whether its occurrence forces the market to re-price.

Catalysts come in several forms. Some are corporate: an acquisition, a partnership, a divestiture, a management change, the launch of a product that meaningfully shifts the revenue mix. Some are cyclical: the trough of a commodity cycle, the inflection of a memory market, the moment a sector emerges from over-supply. Some are narrative: a thematic re-rating that pulls overlooked names into the conversation.

Every Spinel deep-dive thesis will identify the catalyst explicitly, with a stated horizon. If the catalyst materializes and the position does not move, the thesis was wrong about the link between the two, and that will be written. If the catalyst fails to materialize within the horizon, the position will be closed and the work re-examined in writing. The catalyst is the contract.

Asymmetry: The Only Acceptable Shape of a Position

Stock picking is not a game of being right more often than wrong. It is a game of being right by a lot when right, and wrong by a little when wrong. The first part is well understood. The second is where most investors lose.

Every position taken by Spinel must satisfy a minimum asymmetry test before publication. The downside scenario (defined as a credible, examined, written-down bear case) must be limited, ideally bounded by tangible value, replacement cost, or a floor in cash flows that the market has already priced. The upside scenario (defined as the thesis playing out on its stated horizon) must offer at least three times the downside in expected return. Positions that do not clear this bar do not get written about, however interesting the story.

The asymmetry test is what makes contrarian, catalyst-driven investing survivable. Being early is the same as being wrong, except when the downside is already priced in. Then being early is just being patient. The whole discipline collapses without this final filter.

Every thesis begins with a reason to buy and a reason to sell. I write the latter before the position is taken.

The Three Together

Each pillar on its own is insufficient. Contrarian without a catalyst produces a value trap. A catalyst without contrarian positioning produces a crowded trade where the upside is already in the price. Asymmetry has nothing to filter when the first two are absent: the opportunities never arise. The three together produce the only kind of position worth writing about: a mispriced company, with a datable catalyst, bought at a price where being wrong costs little and being right pays a lot.

Every deep-dive note published in these pages will be evaluated against these three pillars. Where a thesis fails the test, the position is not taken. Where it passes, it gets the full treatment. The Weekly format is not bound by these pillars in the same way; its purpose is to develop the broader thinking that informs the work, including analyses of companies that will never become Spinel positions.

What Qualifies as a Spinel Idea

The three pillars describe the philosophy. They do not, on their own, tell you what an idea must look like to clear the bar. The operational test is simpler, and it is written here so that you can hold every published Deep Dive accountable to it.

A Spinel idea must satisfy, in writing, before any position is taken:

A clear market misunderstanding. What does the market believe today, and why is that belief incorrect, incomplete, or stale?

A catalyst that can reasonably occur within twenty-four months. What event, on what horizon, will force the market to update its expectations?

A downside that can be articulated before entry. What is the bear case, what already protects capital if the thesis fails, and what does the market already price in?

An explicit exit framework. At what price does the thesis play out, at what price does it require re-examination, and at what price is it abandoned?

A risk/reward profile of at least three to one in the investor’s favor. How much is being risked, against how much is potentially gained, on the stated horizon?

If I cannot answer those five questions clearly, in writing, I do not write about the idea. The objective is consistency of thinking, not mechanical screening. The consistency is real, and you are invited to check it against every published thesis.

What Comes Next

The first live Spinel thesis will publish within the next four weeks. It will follow the structure that every future deep-dive note will follow: origin of the mispricing, catalyst identified, valuation framework, bear case, exit framework with explicit upside target and invalidation level. The position will be timestamped at publication, entered into the Model Portfolio at the closing price that day, and held accountable to its written framework from that moment forward.

Between deep dives, Spinel Weekly will publish a shorter piece each week, in one of the formats described above.

A few commitments worth stating plainly. I will publish an explicit exit framework with every deep-dive thesis, including an invalidation level. I will review performance publicly each quarter, with no cherry-picking and no rewriting of history. Closed positions, winners and losers, will be discussed with the same prominence as new entries. When a thesis breaks, I will say so in writing, placing the original reasoning and the corrected reasoning side by side.

Spinel is built slowly, on purpose. The discipline of the letter is the discipline of the work it describes.

Subscribe

Spinel Capital Research is published weekly, with a monthly deep-dive note. To receive new notes directly, subscribe below.

Subscribe now

I welcome questions, observations, and pushback on published theses at research@spinelcapital.fr. Material objections will be addressed publicly, with attribution to the sender unless anonymity is requested.

Disclaimer

Spinel Capital Research is an independent editorial publication. It is not investment advice, a recommendation to buy or sell any security, or a solicitation of any kind. Nothing published in these pages is tailored to the financial situation, objectives, or risk tolerance of any individual reader. Readers are solely responsible for their own investment decisions and should consult a regulated financial advisor before acting on any information presented here.

The author may hold personal positions in securities discussed in these notes. Personal positions may differ materially from the Model Portfolio in size, timing, and direction, or may not exist at all. The Model Portfolio is a methodological track record of published theses, not a recommendation to replicate any specific allocation. The author is subject to professional restrictions on personal trading; certain names covered in these pages cannot be held personally and are covered solely on their analytical merits.

Forward-looking statements, including price targets and catalyst expectations, are estimates subject to material uncertainty and may prove incorrect. Securities mentioned are subject to market risk, including the possible loss of principal. Past performance, where referenced, is not indicative of future results.

This publication is distributed from the European Union and is intended for an audience of informed investors. It is not directed at, or intended for distribution to, any person in any jurisdiction where such distribution would be contrary to local law or regulation, including but not limited to the United States, where this publication does not constitute the provision of investment research under SEC Rule 15a-6 or equivalent regulation.

By subscribing, you acknowledge that you have read and understood this disclaimer.