Inference Is the Product Now

Nvidia earned $89 billion from its data center line in one quarter, reported on Aug. 26, 2026, up 117 percent in a year¹. Jensen Huang told investors the same day that compute now counts as revenue¹. CoreWeave showed a smaller number attached to a bigger story. Revenue reached $2.575 billion for the quarter, up 112 percent². Behind it sits a $104 billion book of signed work, plus another $25 billion in the first weeks of the next quarter².

Nebius signed $46 billion in AI cloud deals, led by a five-year, $17.4 billion Microsoft contract³⁴. Crusoe pushed its booked capacity toward five gigawatts, and fresh money valued it near $30 billion⁵⁶. Here is my claim. One inference specialist or neocloud will post a bigger yearly AI line than at least one hyperscaler's own AI business. That crossover happens before the end of 2028.

The number Nvidia set

Huang framed the quarter in unusually direct terms. "AI has reached its inflection point," he said. "It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue. And demand is accelerating"¹. Guidance for the next quarter calls for $108 billion. A decade ago that was a full year of sales for most software companies.

His sharper observation came next, when he described "a golden age of new AI labs and startups, multiple frontier labs scaling in parallel"¹. Read that as a forecast about market structure rather than enthusiasm. Demand is splitting across dozens of labs, and those buyers stop requiring a hyperscaler's balance sheet, which means a neocloud's contract serves them as well as a Microsoft or Google order would.

CoreWeave's backlog

Skeptics spent 2025 calling CoreWeave a one-customer bet financed by debt, and Michael Intrator, its chief executive, heard that charge all year. The second-quarter 2026 filing answers with a customer list. New enterprise names arrived: Bentley Systems, Caterpillar, Grammarly, Isomorphic Labs and Sunday Robotics². Deeper work now runs with Cognition, Databricks, Hudson River Trading, Periodic Labs, Rescale and Runway ML².

Revenue hit $2.575 billion for the quarter, up 112 percent from $1.212 billion a year earlier². Backlog grew to about $104 billion by June 30, with another $25 billion signed within weeks². Intrator named the turn on the earnings call. "CoreWeave reached an important inflection point this quarter as our scale began to translate into expanding operating leverage."² He described the capacity position more bluntly to analysts. "Our near-term capacity remains effectively sold out."²

CoreWeave also claimed the first bring-up and validation of Nvidia's Vera Rubin NVL72 platform, which keeps it inside Nvidia's supply chain as a favored partner². Hold that 112 percent growth at half its pace through 2028, and yearly revenue alone clears $25 billion. Few hyperscalers report an AI line that large.

Nebius and Crusoe

Contract size became the pitch for both firms in 2026. Nebius built its case in April, when signed AI cloud deals reached $46 billion³. Microsoft's five-year, $17.4 billion data center deal leads them, with $7 billion paid upfront⁴. Crusoe built its case in gigawatts instead⁵. Booked capacity pushed toward five gigawatts across its data centers and cloud, and fresh money then valued the firm near $30 billion⁵⁶.

Model ownership sits outside either story, and so does big-name research. Both firms hold deals that top the yearly revenue of most public software makers. Each sells the one thing every lab, startup and enterprise needs whichever model wins. Capacity.

Cerebras and Groq

Scale questions got two answers here. Cerebras chose the public one. Andrew Feldman, co-founder and chief executive, took it public in May 2026, raising $5.5 billion, and shares jumped 108 percent on debut⁷. A harder quarter followed. Cerebras stock fell 10 percent in June, after the firm guided toward a thinner margin in its first report since the listing⁸. Investors read that as a signal. The market still judges it as a chip maker rather than a story.

Groq chose private capital. Jonathan Ross, its founder and chief executive, closed a $350 million Series A at a $3.5 billion valuation, with Nvidia among the backers, then came back for $650 million more⁹¹⁰. Four capital paths, Cerebras public, Groq private, CoreWeave leveraged and Nebius contract-backed, reach one conclusion. Specialist compute firms raise money as fast as any model builder, and investors keep rewarding the specialty over the generalist.

The falling price

Every firm here lives with one hard fact. The token they sell keeps getting cheaper. GPT-4-equivalent output cost $30 per million input tokens in March 2023¹¹. Open-weight models matched that quality for roughly a dime by 2026, a drop near 1,000-fold in three years¹¹. Analysts who track the curve expect prices to keep falling four to tenfold a year through 2027¹¹.

A hyperscaler absorbs that drop across search, apps, retail and cloud storage, where the line item barely registers. Neoclouds meet the same drop with one product to sell, which forces them to run lean and explains a visible pattern. CoreWeave, Nebius and Crusoe keep signing capacity deals ahead of research deals. Gartner sized the prize at $23.3 billion of inference spending in 2026, 55 percent of a $42.276 billion AI market, climbing toward $39 billion of $66.143 billion by 2027¹². Specialists built for this drop, and for this volume, stand ready to take it.

By the numbers

• Nvidia's data center revenue hit $89 billion for the quarter, up 117 percent in a year¹.
• CoreWeave's signed backlog reached $104 billion by June 30, 2026, plus $25 billion more within weeks².
• Nebius anchors a $46 billion deal book with a $17.4 billion, five-year Microsoft agreement³⁴.
• Crusoe runs about five gigawatts of contracted AI capacity across data centers and cloud⁵.
• Cerebras raised $5.5 billion in its May 2026 IPO, then shares rose 108 percent on day one⁷.
• Groq's $350 million Series A valued the Nvidia-backed firm at $3.5 billion, ahead of a $650 million follow-on⁹¹⁰.
• GPT-4-equivalent output cost about 1,000 times more in March 2023 than in 2026¹¹.
• Inference took $23.3 billion of Gartner's 2026 AI-optimized forecast, 55 percent of the total¹².

What to watch

CoreWeave's next filing should show whether the $25 billion in early third-quarter bookings turns into booked revenue on schedule². That is the clearest test of backlog quality in the sector. Cerebras owes investors a margin update on its next call, where a second weak quarter proves the skeptics right and a recovery proves the IPO right⁸. Nebius and Crusoe each have room to announce a hyperscaler-scale contract before 2027 opens, and either move brings the 2028 threshold closer³⁵. Nvidia's $108 billion guidance sets the bar every specialist here now measures against¹. AI Lately will keep a running scoreboard as each company reports.

Sources

1. NVIDIA, "NVIDIA Announces Financial Results for Second Quarter Fiscal 2027," NVIDIA Newsroom, Aug. 26, 2026, https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-second-quarter-fiscal-2027
2. CoreWeave, "CoreWeave Reports Strong Second Quarter 2026 Results," CoreWeave Investor Relations, August 2026, https://investors.coreweave.com/news/news-details/2026/CoreWeave-Reports-Strong-Second-Quarter-2026-Results/default.aspx
3. The Motley Fool, "Nebius Has Landed $46 Billion in AI Cloud Deals. Could This Stock 10X From Here?," The Motley Fool, April 19, 2026, https://www.fool.com/investing/2026/04/19/nebius-has-landed-46-billion-in-ai-cloud-deals-cou/
4. The Energy Mag, "Microsoft Committed $7 Billion Upfront in Nebius AI Deal," The Energy Mag, May 4, 2026, https://theenergymag.com/news/2026-05-04/microsoft-nebius-ai
5. Crusoe, "Crusoe's Contracted AI Infrastructure Capacity Approaches 5 Gigawatts Across Data Centers and Cloud," Crusoe, 2026, https://www.crusoe.ai/resources/newsroom/crusoes-contracted-ai-infrastructure-capacity-approaches-5-gigawatts-across-data-centers-and-cloud
6. Bloomberg, "Crusoe in Talks to Raise $3 Billion in Round That May Triple Firm's Value," Bloomberg, July 2, 2026, https://www.bloomberg.com/news/articles/2026-07-02/crusoe-in-talks-to-raise-3-billion-in-round-that-may-triple-firm-s-value
7. TechCrunch, "Cerebras raises $5.5B, then stock pops 108%, in the first huge tech IPO of 2026," TechCrunch, May 14, 2026, https://techcrunch.com/2026/05/14/cerebras-raises-5-5b-kicking-off-2026s-ipo-season-with-a-bang/
8. CNBC, "Cerebras falls 10% after chipmaker forecasts shrinking margin in first earnings report since IPO," CNBC, June 23, 2026, https://www.cnbc.com/2026/06/23/cerebras-cbrs-q1-earnings-report-2026.html
9. TechFundingNews, "NVIDIA-Backed Groq Raises $350 Million at $3.5 Billion Valuation as AI Inference Race Accelerates," TechFundingNews, 2026, https://techfundingnews.com/nvidia-backed-groq-raises-350m-at-3-5b-as-ai-inference-race-accelerates/
10. Groq Newsroom, "Groq Raises $650M to Scale Its AI Inference Cloud Business," Groq, 2026, https://groq.com/newsroom/groq-raises-usd650m-to-scale-its-ai-inference-cloud-business
11. ValueAddVC, "AI Inference Cost Reduction 2026: Down 95% in Two Years," ValueAddVC, 2026, https://valueaddvc.com/blog/how-ai-inference-costs-have-dropped-95-in-two-years-and-what-happens-next
12. Gartner Newsroom, "Gartner Forecasts Worldwide AI-Optimized IaaS Spending to Grow 96% in 2026," Gartner, Aug. 10, 2026, https://www.gartner.com/en/newsroom/press-releases/2026-08-10-gartner-forecasts-worldwide-artificial-intelligence-optimized-iaas-spending-to-grow-96-percent-in-2026