NextFin News - Nvidia reported quarterly revenue of $96.2 billion, more than double the year-earlier figure, while Salesforce's shares jumped about 11% after a profit print lifted by a multibillion-dollar gain on its Anthropic stake. The twin results on August 26 frame the same question from opposite ends of the artificial-intelligence trade: the infrastructure layer is still accelerating, but the software layer now has to prove it can pay the bill.
The Numbers: A Buildout That Is Still Speeding Up
Nvidia's second quarter of fiscal 2027, ended July 26, delivered revenue of $96.2 billion, up 18% from the prior quarter and 106% from a year earlier. Data Center revenue — the segment that matters — reached $89.0 billion, up 117% year over year and roughly 93% of the total. Both GAAP and non-GAAP gross margin came in at 75.0%, and diluted earnings per share were $2.46 on a GAAP basis and $2.22 adjusted. That beat the consensus on the street of $2.10 per share and $92.17 billion in revenue, according to LSEG.
The more telling number is the guide. Nvidia expects third-quarter revenue of $108.0 billion, plus or minus 2% — a figure that implies the company will add more revenue in a single quarter this fall than it generated across all of fiscal 2024. Gross margin is guided to 74.0%, plus or minus 50 basis points, a modest step down that reflects the mix shift toward new platforms rather than any loss of pricing power. For the full fiscal year, the company expects tax rates between 16% and 18%.
Hours earlier, Salesforce reported its own fiscal second quarter, ended July 31: revenue of $11.3 billion, up 11% year over year, with subscription and support revenue up 12%. GAAP diluted earnings per share were $4.29, up 119%, and non-GAAP earnings were $5.90, up 103%. Current remaining performance obligation — a forward-looking bookings proxy — rose 14% to $33.5 billion, ahead of the $33.22 billion analysts had expected. Free cash flow climbed 81% to $1.1 billion.
The two prints landed on the same afternoon, and the market read them as a single story. Nvidia shares rose about 4% after the close and extended gains into the next session, while Salesforce, down more than 25% for the year through mid-August after falling to a multiyear low of $146.32 in June, reclaimed double-digit percentage ground. One company sells the picks and shovels; the other was supposed to be the casualty of the automation those picks and shovels enable. On this evidence, both are winning.
There is a divergence worth pricing in. Nvidia's growth rate is accelerating — 106% year over year, up from 85% in the prior quarter — while Salesforce's is steady at 11%. The market rewarded both, which means investors are treating Salesforce's quarter not as a growth story but as a survival story. The thesis that mattered was not "how fast is revenue growing" but "is AI killing the CRM franchise." The answer, for now, is no.
Why the Software Layer Matters More Than the Chip Count
The easy read of Nvidia's quarter is that demand for AI accelerators remains insatiable. The harder, more important read is what Salesforce represents: the first real test of whether AI spending is generating revenue on the other side of the transaction.
For most of 2026, the market priced enterprise software as a victim of AI. The logic was clean and brutal: if AI agents can resolve customer cases, write code, and draft marketing copy, then the seats Salesforce sells become expendable. The stock's decline to $146.32 was that thesis expressed in price. The August quarter pushes back.
Salesforce now reports Agentforce annual recurring revenue above $1.5 billion, up more than 240% year over year. Combined with Data 360, AI and data ARR reached nearly $3.9 billion, up more than 210%. Usage is scaling faster than revenue: the company said it has delivered 7.0 billion "Agentic Work Units" to date, with 3.2 billion in the quarter alone, up 97% from the prior three months. Data 360 ingested 104 trillion records in the quarter, up 355% year over year, including 82 trillion via Zero Copy, up 731%.
"We just delivered one of our best quarters ever, outperforming across every key metric," said Marc Benioff, Salesforce's chair and chief executive. "AI is delivering value across every layer of our platform."
The strategic subtext is the Anthropic relationship. Salesforce built a stake in the AI lab through repeated rounds starting with Series C in early 2023 — roughly 1% of the company, with total capital committed above $300 million and a stake valued at about $5 billion by mid-year. The August quarter included a $2.6 billion gain on strategic investments tied to that position, which lifted both GAAP and adjusted profit. Separately, Benioff has said the company is on track to spend about $300 million on Anthropic tokens in 2026, mostly for coding and internal productivity.
That is the pivot in one frame: Salesforce is no longer just a buyer of AI. It is an investor in the model layer, a consumer of the tokens, and — critically — a monetizer of the output inside the enterprise workflow. If AI were purely cannibalistic, none of those three roles would be growing revenue. The quarter suggests AI is additive to the CRM relationship, not a substitute for it.
There is a caveat on earnings quality. Of the $5.90 in non-GAAP earnings per share, roughly $2.53 came from the $2.6 billion investment gain rather than operations. Strip it out and operating EPS was closer to $3.37. The market accepted the headline anyway, which is itself a signal: investors are treating the Anthropic stake as a strategic asset on the balance sheet, not a one-time windfall to be discounted.
"NNAOV growth is the strongest it's been in four years, keeping us on track for second-half organic revenue reacceleration," said Robin Washington, Salesforce's president and chief financial and operating officer.
The company raised full-year revenue guidance to $46.1 billion to $46.4 billion, up 11% to 12% year over year, from a prior range of $45.9 billion to $46.2 billion. Third-quarter revenue is guided to $11.42 billion to $11.5 billion. Management also said it expects to close the pending acquisitions of Contentful and Fin in the coming weeks, and both are now incorporated into guidance.
The $500 Billion Question: Who Actually Pays for the AI Factory
Nvidia's quarter contained a development that matters more than any single revenue line. In August the company announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent compute-financing platforms intended to mobilize more than $500 billion of third-party capital for AI infrastructure.
The mechanism is the story. Until now, the buyers of Nvidia's $89 billion quarterly data-center output have been almost entirely the hyperscalers and a small set of well-funded frontier labs — entities with the cash flow or the venture funding to write nine-figure checks upfront. The financing platforms are designed to turn AI factories into an investable asset class, much like commercial real estate or toll roads, so that pension funds, insurers and sovereign capital can own the infrastructure and Nvidia can sell against usage-linked, long-duration revenue.
This is financialization of AI capex, and it changes the shape of the cycle. If it works, the funding base for the buildout widens from a handful of tech balance sheets to the global pool of institutional capital. Nvidia's own language — "long-duration usage-linked revenue" — describes a business model that looks less like a chip cycle and more like an infrastructure annuity.
The structure also reveals a second-order shift inside Nvidia itself. The company is moving from selling boxes to underwriting the capital stack that buys the boxes. That gives it leverage over a wider customer base — sovereign builds, enterprises and labs that cannot fund a full rack upfront — but it also exposes Nvidia to the credit cycle. A chip company does not usually worry about the debt service of its customers. An infrastructure financier does.
It also broadens the risk. The same leverage and duration that make infrastructure investable make it fragile when utilization falls short. The terminal payer in every structure is whoever rents the compute — which loops the whole chain back to the Salesforce question: is enterprise software actually generating enough incremental revenue to service the infrastructure that serves it?
The peer set is watching. Alphabet, Microsoft and other large investors in Anthropic have flagged similar investment gains, which means the Salesforce print is not an isolated accounting event. It is the first visible line of a broader repricing: the model layer is being marked to market on the balance sheets of its corporate backers, and those marks are now flowing into GAAP earnings.
The Bear Case: Circular Capex and the Monetization Gap
The strongest argument against the bull case is not that the chips don't work. It is that the revenue is circular. Hyperscalers book Nvidia sales today against AI application revenue that may arrive years from now, if ever. A widely cited analysis from Sequoia Capital asked precisely this: whether the hundreds of billions being spent on AI infrastructure can be justified by the software revenue it is likely to produce. The concern gained fresh attention in 2026 as reports surfaced of "circular AI financing" — customers funding AI purchases with capital raised against the promise of AI revenue.
Nvidia's financing partnerships could extend the circle rather than close it. If asset managers fund data centers whose tenants are AI labs whose only revenue is the compute subsidy, then the $500 billion is not new demand — it is new credit layered on top of the same speculative premise. The difference between a structural buildout and a credit-fueled bubble is whether the end customer pays. And the end customer is the enterprise software buyer Salesforce serves.
There is also a concentration risk inside Nvidia's own numbers. Data Center at 93% of revenue is a success story and a vulnerability in the same line. The company is not assuming any data-center compute revenue from China in its outlook, and the guide assumes the Vera Rubin ramp lands without supply disruption. Any stumble in the handful of customers absorbing $89 billion a quarter shows up immediately. Networking revenue reached a record and Spectrum-X Ethernet grew 2.6 times year over year, but that growth is still tethered to the same small group of hyperscale buyers.
The bull answer is that monetization is arriving, just on a lag — and that the financing layer exists precisely to bridge that lag. Salesforce's 240% Agentforce growth and 97% sequential jump in agentic work units are the earliest hard evidence that the applications layer is catching up to the infrastructure layer. The bear answer is that usage growing faster than revenue is exactly what a subsidy looks like before the bill comes due.
Both sides can point to the same data. That is why the next two quarters matter more than this one.
What to Watch: The Signals That Break the Thesis
The structural call here is that AI compute is becoming a financed utility — a durable shift in how technology capital is formed. The cyclical call is that the current 100%-plus growth rates and multiple expansion are a super-wave that will normalize. Holding both views at once is not a hedge; it is the only way to read the data honestly.
Three signals separate the two. First, Nvidia's data-center growth rate: if it decelerates below 50% year over year for two consecutive quarters while hyperscaler capital-expenditure guidance holds flat or falls, the acceleration thesis is broken. Second, Salesforce's monetization ratio: if Agentforce ARR growth slows below 100% year over year for two quarters while agentic work units keep doubling, adoption is outrunning willingness to pay. Third, the financing platforms themselves: memorandums of understanding are not capital deployed. Signed, funded vehicles — not announcements — are the proof.
For the near term, momentum favors the infrastructure layer. Nvidia's $108 billion guide implies the cycle has further to run, and Salesforce's rerating is the market's first serious reconsideration of the "AI kills SaaS" thesis. Over the medium term, the ceiling is set by the financing layer: how much of the $500 billion converts into deployed capacity, and at what cost of capital. Over the long term, the winners are likely to be the companies that own the customer workflow as much as those that sell the silicon — because in an era of financed compute, the party that controls the invoice has the leverage.
Scenarios branch from there. In the base case, Nvidia grows into the guide, Salesforce holds 11% to 12% revenue growth, and AI capex compounds at 30% to 40% annually. In the upside case, the financing platforms close and bring genuinely new buyers, enterprise AI spend reaccelerates, and Nvidia's data-center growth stays above 80%. In the downside case, circularity concerns trigger a capex pause, hyperscalers trim guidance, and the software narrative reverts to disruption fear.
This rally is being underwritten not by better chips but by a new balance sheet. The test ahead is whether enterprise revenue can service it.
Explore more exclusive insights at nextfin.ai.
