Maintaining Outperform: Headline Beat, RPO $138B, and an FY26 Raise to >$67B Confirm the Bookings-to-Revenue Cascade Is Delivering

Rating Action: Maintaining Outperform

We are maintaining Oracle Corporation at Outperform following the Q4 FY2025 print. The Q4 release confirmed, on every operative metric, the cascade we underwrote in March: bookings continued to translate into a step-up in the multi-year revenue framing, the cloud infrastructure acceleration carried into the year-end print, and management raised forward guidance across every horizon (Q1, FY26, and the FY27/FY29 long-range targets). The capex absorption math worsened in-quarter and the FY25 FCF print was the cleanest negative since the late 1990s, but the offset is structural: capacity, not capital and not demand, is the binding constraint, and Oracle is investing through the trough in line with a revenue trajectory that is now on the record.

Arc Walk

• Q2 FY2025 (Initiating at Hold, constructive bias) — Clean operational beat (RPO $97B +50% cc, OCI +52%) but a penny EPS miss against $4B-quarter capex absorption that drove Q2 FCF to -$2.7B. We initiated at Hold on three reservations: capex absorption unproven on a multi-quarter horizon, AI customer concentration asymmetric (Stargate/OpenAI as the largest single dependency), and multi-cloud revenue still small in dollar terms. The upgrade trigger list was specific: RPO at or above $110B, capex tracking at or below the doubled-FY24 framing without further upward revision, multi-cloud exiting FY25 on a path to $1B+ for FY26.
• Q3 FY2025 (Upgrading from Hold to Outperform) — The bookings step-function: RPO blew past the trigger to $130B (+63% cc) on a $48B in-quarter add, cloud RPO crossed 80% of total at +90% growth, multi-cloud grew 200% in three months across 18 live and 40 in-deployment regions, customer concentration diversified via the AMD MI355x deal and four cloud-security wins, and Q3 OCF funded Q3 capex 1:1. Management hardened the FY26 +15% framing and lifted FY27 to ~+20%. We upgraded.
• Q4 FY2025 (Today — Maintaining Outperform) — The cascade is delivering. Headline beat both lines (rev +11%, EPS $1.70 vs. ~$1.64 consensus), RPO continued higher to $138B (+41% USD), the FY26 framing was raised to >$67B (+16% cc) and now carries an explicit >100% RPO growth guide that pre-commits to another year of mega-bookings. Cloud infrastructure is now guided >70% in FY26 (vs. 51% in FY25), total cloud >40% (vs. 24%). Multi-cloud regions stepped from 18+40 to 23+47. The FY25 FCF print of -$0.4B and the FY26 capex bump to >$25B are the cost of the trajectory, not a refutation of it. The thesis remains intact; the magnitude has only grown.

What gets us back to Hold or to Underperform from here: (a) RPO conversion to revenue lags the path implied by the FY26 +16% target — specifically, if Q1 or Q2 FY26 prints come in below the +11-13% cc Q1 guide; (b) capex revised upward beyond the new $25B+ framing without commensurate bookings or revenue progression (a $30B+ capex year on its own would not be disqualifying, but it would be on the watch list); (c) any signal of customer-commitment slippage on the announced AMD or Stargate-related deals as Stargate moves from formation to revenue-generating contracts; (d) FY26 FCF more negative than -$5B without a credible bridge to FY27 FCF inflection; (e) gross margin compression at the OCI segment level that contradicts the “Gen 2 is structurally cheaper” thesis.

Results vs. Consensus

Q4 was the cleanest beat on both top and bottom lines we have seen from Oracle in the recent arc, and it came against an unusually wide range of sell-side modeling because the sequential RPO step-up against Q3’s $48B was always going to be the framing risk. The beat happened in the headline; the forward guide is what made the print decisive.

Quality of the Print

• Revenue: A clean, broad-based beat. The 11% USD growth print is the first double-digit revenue growth quarter Oracle has produced since the cloud transition began — Catz called this out specifically in the prepared remarks (“In Q4, we hit double-digit revenue growth and it’s only going up from here”). The beat was distributed across cloud (driven by IaaS +52%), strategic SaaS (+20% on annualized basis), and even software license (+8% — a notable swing from the 8% decline in Q3, with management framing it as “BYOL” activity ahead of cloud migration). FY25 total revenue closed at $57.4B (+9%).
• EPS: $1.70 is a six-cent beat against the high end of consensus and was achieved with a 70-bp tax-rate headwind. The implied operational EPS at the guided 19% tax rate would have been closer to $1.72 — a clean, durable beat with no FX or one-time helps. FY25 non-GAAP EPS of $6.00 is the first $6 print in Oracle’s history and represents +9% growth, with the FY26 algorithm expected to extend the EPS growth alongside revenue acceleration.
• RPO: $138B is +$8B sequentially and +41% YoY (USD). The smaller sequential step versus Q3’s $48B was the easy bear-case angle pre-print, but two structural offsets matter: (1) Catz pre-committed FY26 RPO growth of >100%, which mathematically requires a year of mega-bookings; (2) cloud RPO is now ~80% of total at +56% growth, so the mix continues to harden the multi-year revenue trajectory. Approximately 33% of RPO is expected to convert to revenue over the next twelve months — ~$45B of NTM cloud-derived revenue contribution.
• OCI & Cloud: Q4 IaaS of $3.0B (+52%) is the highest absolute IaaS quarter in Oracle’s history and continues to grow at a multiple of every named hyperscaler peer’s growth rate at scale. OCI consumption was +62%. Cloud database services grew 31% with annualized revenue of $2.6B; Autonomous Database consumption was +47% on top of +27% (the two-year compound is now in the same territory as the OCI growth profile). Total cloud (SaaS+IaaS) of $6.7B grew 27% — an acceleration off the +25% Q3 print.
• Cash flow: The cleanly-bearish line of the print. Q4 capex of $9.1B blew past the implied run-rate from the prior $16B FY25 framing — meaning Catz raised the FY25 capex envelope twice in two quarters and ended at $21.2B. Q4 FCF of -$2.9B and FY25 FCF of -$0.4B mean Oracle finished its strongest revenue-acceleration year as an FCF-negative business. The right read is not “the cash engine is broken” — OCF grew 12% to $20.8B for the year — but rather “the capacity build-out is consuming the entire OCF base and a small slice on top.” FY26 capex of >$25B (with Catz signaling it could be higher) extends the FCF trough by another year. We model FY26 FCF in the -$2B to -$5B range pending capex pacing.

Segment Performance & Business Drivers

Oracle Cloud Infrastructure (OCI) — the central story, again

OCI revenue of $3.0B grew 52% YoY (versus 51% in Q3 and 49% in the prior-year comparable). Annualized run-rate of nearly $12B is now nearly half of the total cloud business. OCI consumption (the closer-to-real-time demand signal) accelerated to +62% (versus +57% in Q3). The FY26 guide of >70% OCI growth implies an absolute FY26 OCI revenue contribution of approximately $17B+ — an additional $7B of revenue versus FY25, exceeding the entire FY24 OCI base in incremental terms.

The structural read on the OCI line is that the +50% growth rate is now eight quarters old and accelerating, which is not what large-base growth normally does. Management is attributing this to the Gen 2 architecture’s capacity utilization advantages: capacity comes online and fills immediately on backlog, with deployment cadence the binding constraint, not demand creation. The FY26 acceleration to >70% is consistent with capacity flowing on-stream and existing RPO converting in the second half of the fiscal year.

Multi-Cloud Database (Azure / Google / AWS)

The deployment cadence we wanted continued. Multi-cloud database is now live in 23 regions across the three hyperscaler partners, with 47 more planned — up from 18 live and 40 planned at the end of Q3. Cloud database services grew 31% with annualized revenue of $2.6B; Autonomous Database consumption was +47% (on top of +27% YoY). Larry Ellison’s long arithmetic on the database-to-cloud migration on the call — that 10% of database support revenue moving to cloud equates to ~$50B of incremental cloud revenue once associated compute and networking are layered — is the directional framing for why the database business sits at the structural center of the bull case, not OCI alone.

The 23-region milestone is meaningful because it is the first time the multi-cloud footprint has materially closed the geographic-coverage gap with the partner hyperscalers in their own regions. Oracle no longer needs to win a customer who is “already on Azure” by getting them to leave Azure; the customer can stay on Azure and consume the Oracle database in the partner’s region. That is a fundamentally different go-to-market motion than three quarters ago.

Strategic SaaS & Applications

Strategic back-office SaaS applications now have annualized revenue of $9.3B (+20%), accelerating from the $8.6B / +8% framing we saw in Q3. The acceleration is the inflection Catz teased on prior calls about Fusion and NetSuite ramp on top of embedded AI agents (now 100+ agents announced). Total SaaS revenue of $3.7B (+11%) continues to look pedestrian on the headline but masks the Strategic-vs-non-Strategic split — the Strategic mix is the operational story; the legacy/non-Strategic SaaS line is run-off. Application subscription revenues including support of $5.0B grew 8%. Catz’s framing that “cloud applications growth rate will accelerate this coming year” is the under-appreciated leg of the FY26 algorithm.

License & License Support

Software license revenue of $2.0B (+8%) was the surprise in the print and the segment pre-call modeling had broadly expected to decline. The +8% beat reflects what Catz framed as “BYOL” activity — customers licensing software with intent to deploy in cloud via bring-your-own-license. This is a structural counter-signal to the bear case that on-premise license revenue is in secular decline; what is happening instead is that customers are licensing in advance of cloud migration, which extends the legacy revenue base while also pre-committing the cloud cascade. Database subscription revenues grew 7%; total cloud services and license support of $44B (FY25) was +12% on the year and represents 77% of total revenue.

Key Topics & Management Commentary

The FY26 Algorithm Reset

The FY26 guide was the centerpiece of the call. Every line of the multi-year framing was raised: revenue from $66B to >$67B, total cloud from a previously implicit ~+30% to >+40%, OCI from >+50% (the FY25 print) to >+70%, and RPO from no explicit growth target to >100%.

“So as a result of the strength in our cloud applications and infrastructure including database services, we are raising our revenue guidance for fiscal year 2026 to over $67 billion, up 16% for the year… I expect that total cloud revenue will grow over 40% in constant currency, up from 24% in FY 2025. I expect the cloud infrastructure revenue will grow over 70% up from 51% in FY ‘25… RPO is likely to grow more than 100% in fiscal 2026.” — Safra Catz, CEO

The FY26 +16% revenue print, if delivered, would be Oracle’s highest annual growth rate since the 1990s. Catz also reaffirmed she expects to exceed the prior FY27 target and is “more confident” in the FY29 long-range framing — with a more comprehensive long-range update flagged for Oracle Cloud World in October 2025.

Capex Re-Baseline & the Demand Backdrop

The capex framing was the most analytically charged moment of the call. FY25 ended at $21.2B (versus $16B framing exiting Q3), Q4 alone was $9.1B (versus implied run-rate of $4-5B), and FY26 is now >$25B with Catz adding it “may turn out to be understated.” Larry Ellison contributed the most evocative datapoint of the call on demand:

“We recently got an order that said we’ll take all the capacity you have wherever it is. This could be in Europe, could be in Asia, we’ll just take everything… The demand is astronomical… I’ve never seen anything remotely like this. People are calling up and asking us, please, can you find us more capacity? We’ll take it wherever.” — Lawrence Ellison, Chairman & CTO

Catz reframed the capex spend as demand-anchored: “We don’t order. We don’t build. Unless we’ve got orders for our capacity to be built out.” This is a meaningfully different framing from the speculative-capex risk that has overhung the broader hyperscaler trade in 2025: at Oracle, capacity follows contracted bookings, with the conversion timing being the operational variable. The FCF cost is real; the capital allocation discipline is clearer than the absolute number suggests.

Stargate & OpenAI Partnership Status

Catz directly addressed the question on Stargate that has been the most-modeled forward variable since Q2:

“The reality is that Stargate is still in formation. There are a lot of partnerships we are in the middle of right now. That are all part of this enormous growth rate… As Stargate forms, that will contribute into all of this. But, some of our partners many of our partners some of them will be in Stargate. Some are outside of Stargate.” — Safra Catz, CEO

Larry Ellison closed the Stargate exchange with a striking conditional: “If Stargate turns out to be everything is advertised then we’ve understated our RPO growth.” The translation: the >100% FY26 RPO growth guide is on a baseline that excludes Stargate as a fully-formed contract. Stargate, if it lands as advertised, is upside to a guide that is already aggressive. This is materially constructive language that did not exist three months ago.

The TikTok/ByteDance Disclosure

The new customer disclosure of the quarter was Larry Ellison’s explicit reference to a “gigantic contract from TikTok” and ByteDance moving its infrastructure to Oracle Cloud:

“We got a gigantic contract from TikTok that would have been unprecedented except for all the other gigantic contracts we’re also been getting… This next generation of companies like ByteDance, TikTok, which obviously we do business with them, T Moves and other but Uber, I mean, there are just lots of these companies. A bunch of the security companies have moved to the Oracle Cloud.” — Lawrence Ellison, Chairman & CTO

The point Ellison was emphasizing is that “huge contracts that have nothing to do with even AI” are now part of the bookings mix. The customer concentration broadening we cited as condition #4 of our March upgrade has continued: TikTok/ByteDance, Uber, and the security cohort all sit alongside the AI-training customers in the new RPO base. Stargate is no longer the marginal driver of incremental bookings; it is one input among many.

Application Suite & AI Agents

Larry Ellison gave one of his most extended substantive answers of recent calls on the integrated application suite vs. point-vendor competitive dynamic. The framing — that customers are increasingly preferring an integrated suite of AI-agent-based applications across ERP, EPM, supply chain, manufacturing, HR, and customer engagement, rather than knitting together five point-vendors — is a long-running thesis that is now showing up in win rates. Catz noted Strategic SaaS at $9.3B annualized +20% and explicitly forecast acceleration in cloud applications growth in FY26.

Analyst Q&A — What Was Pressed and What Was Defended

The Q&A was unusually short — six analyst questions versus the typical eight-to-ten — reflecting both the clarity of the prepared remarks and the relatively unambiguous nature of the print. Substantive lines of questioning were focused on (a) the AI business durability and breakdown, (b) whether Stargate is in the FY26 growth and RPO numbers, (c) capex composition and trajectory, (d) cloud database migration economics, and (e) the application-suite competitive position.

Mark Moerdler (Bernstein Research) opened with a direct question on the durability and profitability of the AI business and asked for any color or breakouts. Larry Ellison’s response was the most thesis-relevant exchange of the call: the value-add of Oracle’s position is not the AI training capacity itself but the database-to-AI-model integration that no other vendor offers at scale. The framing — that Oracle has “most of the world’s valuable data” in Oracle databases, that the latest 23ai is a vector database, and that Oracle uniquely enables enterprises to use their own data with major LLMs (Grok, ChatGPT, Llama, Gemini) inside the Oracle Cloud — is the structural inferencing-applied-to-installed-base argument from Q3, now framed as the durability answer.

John DiFucci (Guggenheim) asked the central forward-modeling question: is Stargate part of the FY26 +70% IaaS growth and the FY26 +100% RPO growth? Catz’s answer was deliberately precise: Stargate is in formation, partnerships are in flight, those will contribute to growth, but the explicit FY26 numbers do not require Stargate to fully materialize. Ellison added the upside-risk framing: if Stargate fully forms, “we’ve understated our RPO growth.” This was the cleanest statement we have yet had on the Stargate-as-incremental-upside thesis — the FY26 +100% RPO target is expected to clear without Stargate as the swing factor.

Ben Reitzis (Melius Research) drilled into capex composition and the trajectory beyond FY26. Catz’s response is the disclosure with the cleanest analytical content in the print: Oracle’s capex is “usually about equipment,” not buildings (buildings are leased from third-party builders that charge rent once construction completes). The Q4 step-up in capex was equipment-led, with management taking advantage of available supply (“we had an opportunity to buy up”). The implication for the cash-flow walk is that the equipment-heavy capex profile means depreciation and revenue conversion track more tightly than they would for a real-estate-heavy build-out. Ellison added that GPU supply is not the constraint at Oracle (“No. No. Not right now”), pivoting the binding-constraint framing back to power and high-speed networking.

Siti Panigrahi (Mizuho) raised the on-premise database migration economics. Catz reaffirmed the broad-based momentum (license up, support solid, autonomous consumption up, multi-cloud taking up all available capacity) and reframed the database business as “on fire.” Larry Ellison provided the long-form arithmetic: 10% of $100B+ on-premise database support converting to cloud equals at least $50B of incremental cloud revenue once associated compute and networking layers are included. The translation: even a slow migration cadence has the potential to drive a $50B+ incremental cloud opportunity from the existing database installed base alone.

Raimo Lenschow (Barclays) raised the application business specifically — the segment that has been overshadowed by OCI in the bull-case framing — noting Fusion at +22% growth. Catz’s answer separated the Strategic SaaS portfolio (going “gangbusters,” +20% annualized) from the legacy/advertising businesses (now lapping). Ellison gave the most extended substantive answer on the integrated-suite-vs-point-vendor positioning, framing the win-rate momentum in vertical applications (Oracle Health, Oracle Banking, Oracle Hospitality) as a structural moat that compounds across more enterprise customers each quarter.

Brad Zelnick (Deutsche Bank) closed with the strategic-positioning question on Oracle’s full-stack architecture from silicon to applications and why that integration matters in the new era. Larry Ellison gave the historical-arc answer: each layer of the stack informs and improves the others — building applications informed the database; building the database enabled the cloud; the autonomous database makes the cloud more reliable, more secure, and lower-cost-to-operate. The framing positions integration as a structural advantage that compounds over time and is not replicable by point-vendor competitors without an equivalent multi-decade engineering investment.

What They’re NOT Saying

• FY26 FCF magnitude. Catz did not provide an explicit FY26 FCF framing. With FY26 capex now at >$25B (and Catz signaling it could go higher) against an FY25 OCF base of $20.8B that is growing at +12%, the implied FY26 OCF base is roughly $23B. The implied FY26 FCF is therefore in the negative $2B to negative $5B range depending on capex pacing — a second consecutive negative annual FCF print after FY25’s -$0.4B. Management is clearly choosing to underwrite the demand cascade through the FCF trough; readers should size positions assuming negative FY26 FCF.
• Sequential RPO step-down dynamics. Q3 added $48B to RPO; Q4 added $8B. Management did not explicitly address why the sequential cadence cooled. Some of this is mechanical (Q3 had a particularly large bookings concentration including AMD MI355x); some may reflect the fact that Q4 conversion to revenue from the existing RPO is now larger in absolute terms (revenue +11% off a higher base subtracts more from the gross-additions to net-RPO). The FY26 +100% RPO guide implies the next four quarters will average $20-30B of net-additions per quarter — a cadence that requires Stargate, AMD additions, or a continued mega-deal rhythm.
• OCI segment-level gross margin. As in Q3, Oracle does not break out OCI segment-level gross margin at the granularity that would let third parties model it independently. Management asserts OCI gross margins are trending higher, but the consolidated cloud services and license support gross profit dollars are growing slower than revenue cc, implying the multi-cloud and SaaS mix is masking some of the OCI margin signal.
• Power capacity by region (still not disclosed). Catz disclosed in Q3 that power capacity would double in CY25 and triple by end-FY26. The Q4 call did not provide an updated geographic breakdown. Given Larry’s “we’ll take it in Malaysia” vignette, the constraint is meaningfully geographic. We would have preferred a North America vs. Europe vs. Asia breakdown, particularly with $25B+ of FY26 capex flowing.
• Specific Stargate timing. “Stargate is still in formation” is the same framing we received in Q3’s “fairly soon”. Three months later, it is still in formation. We do not yet have a quarter or fiscal year for the first revenue dollar from Stargate, even as Catz committed to the FY26 +100% RPO guide that does not require it. The opacity is uncomfortable but the framing — Stargate as upside, not baseline — is now well-established.

Market Reaction

Shares traded materially higher in extended hours on June 11 following the print, with the after-hours move on the order of +6-8% and a continuation of the strength into the June 12 regular session, where Oracle traded at fresh all-time highs as the market priced the FY26 algorithm reset and the cloud infrastructure +70% guide. The print was the cleanest direct confirmation in the recent arc that the Q3 bookings step-up was not a one-quarter outlier but the start of a multi-year acceleration. The trading reaction reflects that interpretation: not a relief rally on the headline, but a re-rating on the forward framing.

The medium-term valuation question is whether the multiple now extends or holds as the cascade actually delivers. Oracle’s equity has re-rated sharply through the upgrade arc, and the FY26 +16% revenue and +70% OCI guides are now embedded in consensus. The next two prints (Q1 FY26 and Q2 FY26) need to track the Q1 +11-13% cc revenue guide and continue to add to RPO at a $20B+ quarterly pace to validate the FY26 trajectory. Misses on either front would be received harshly given the multiple. Beats — particularly on RPO — would extend the re-rating further.

Street Perspective

The Street perspective heading into Q4 was constructive but cautious about the sequential RPO cadence after the historic Q3 print. The bull case being made coming in was that the headline would beat (revenue and EPS both in the slight-beat zone), that RPO would hold above $130B with a smaller sequential addition, and that the FY26 framing would be reaffirmed without further raise. The actual print delivered the headline beat as expected, held RPO above $130B with a modest sequential add, and meaningfully raised the FY26 framing — the third leg was the surprise.

The bull case being made on the Street post-print is that Oracle has now decisively transitioned to a credible #4 hyperscaler with structural AI inferencing and multi-cloud differentiation, and that the FY26 +16% revenue and +70% OCI growth are now achievable rather than aspirational. The bear case being made is that the FCF profile is structurally worse than the bull case acknowledges — FY25 -$0.4B, FY26 likely negative again, FY27 still capex-heavy — and that the equity is priced as if the FCF inflection is closer than the math suggests. Both views have analytical support; we sit on the bull side because the bookings-and-capacity story has tightened in every quarter of the arc, and the FCF trough — while painful — is now characterizable in dollars and quarters rather than open-ended.

Model Implications

• Revenue: FY25 closed at $57.4B (+9% USD). FY26 explicitly raised to >$67B (+16% cc) — modeling $67.5B mid-case with skew to upside if Stargate flows. FY27 implicitly higher than the prior +20% framing per Catz’s “exceed the FY27 target” comment — modeling FY27 ~$80B (+19-20%). FY29 reaffirmed and now “more confident”, with a more fulsome update due at Oracle Cloud World in October.
• Cloud revenue mix: Total cloud (SaaS+IaaS) +>40% in FY26 implies $34B+ of total cloud revenue. OCI alone >+70% implies ~$17B+ of OCI revenue in FY26 — meaning OCI standalone exceeds the entire cloud business of just two years ago.
• Operating margin: Q4 operating income +7% lagged Q4 revenue +11%, reflecting the depreciation step-up from accelerating capex. Modeling FY26 non-GAAP op margin in the 41-43% range — structurally lower than the 44% Q3 print reflected, because the depreciation cycle from the $25B+ capex flows hits before the corresponding revenue conversion fully ramps. The operating-leverage thesis remains intact on a multi-year horizon, but FY26 is the absorption year.
• Capex: FY25 $21.2B; FY26 >$25B (potentially $28-30B). The capex envelope is the single most important variable in the FY26 cash-flow walk. We model FY26 capex at $27B mid-case.
• FCF profile: FY25 -$0.4B. FY26 -$2B to -$5B (modeled mid-case -$3B given $23B implied OCF and $27B capex). FY27 inflection back toward positive in the $3-7B range as the capex-to-revenue lag closes. FY28+ stepping toward $15-20B annualized as the investment cycle moderates. The FCF trough is now multi-year rather than single-year; the offset is that the bookings backlog and capacity commitments make the eventual inflection clearer than at any prior point.
• Capital return: Q4 buyback was again modest at $150M (~1M shares); FY25 buyback was modest by historical standards. Dividend held at $0.50/quarter; FY25 dividend payments totaled $4.7B. Capital allocation is appropriately tilted toward capacity investment and the dividend over near-term buybacks, consistent with the bookings opportunity and the FCF profile.

Thesis Scorecard

Bottom Line

Oracle Q4 FY2025 was the conversion print our March upgrade required. The headline beat both lines for the first time in the recent arc; the cloud and OCI segments accelerated; multi-cloud continued its deployment cadence; the FY26 framing was raised across every line of the algorithm; and the FY27/FY29 long-range trajectory was upgraded twice (from “reaffirmed” to “exceed” and from “reaffirmed” to “more confident”). The FY26 +100% RPO target pre-commits to a backlog inflection without precedent in Oracle’s history, and crucially, does not require Stargate to fully form to clear — Stargate is now framed explicitly as upside.

The cost of the trajectory is real: FY25 ended at -$0.4B FCF (the first negative annual FCF print since the late 1990s), Q4 capex of $9.1B re-broke the OCF/capex coverage we celebrated last quarter, and FY26 capex at >$25B all-but-guarantees a second consecutive negative-FCF year. The offsetting framing — that capex follows contracted bookings, that GPU supply is not the constraint, that power capacity is on a doubling-then-tripling trajectory, and that demand is “astronomical” with customers offering to take capacity wherever it can be delivered — makes the FCF trough underwritable on the demand math, but readers should size positions assuming that FY26 will be a second consecutive negative-FCF year.

The thesis from our March upgrade is now stronger across every operative dimension. The bookings cascade is delivering. The forward acceleration is on the record, in management’s words, with explicit triple-digit RPO and double-digit revenue guides. The trough is dollarized and bracketed. Maintaining Outperform.