NVDA's $108B Guide Is the Only Number That Matters
NVIDIA guided Q3 to $108 billion — the largest data center revenue forecast in semiconductor history. Here's what it means for hyperscaler capex and your portfolio.
Key Points
- NVIDIA guided Q3 FY27 revenue to $108 billion, ±2%, with data center alone at $89 billion in Q2 — both figures are the largest in the company's history.
- Every major hyperscaler's AI infrastructure budget now flows directly through NVIDIA's order book, making the guide a proxy for the entire sector's capex trajectory.
- Traders should watch Broadcom's custom ASIC commentary this week as the cleanest live signal on whether hyperscalers are beginning to route spending away from NVDA silicon.
NVIDIA's Q2 FY27 print landed at $96 billion in total revenue, with data center contributing $89 billion — and then the company guided Q3 to $108 billion. That sequential jump of $12 billion in a single quarter is not a beat, it is a structural reset of what AI infrastructure spending looks like at scale. NVDA closed Monday at $227.98, up 15.72% over the past month and 22.39% year-to-date, and the stock's trajectory since the print makes clear that the market read the guide correctly.
The Number Behind the Number
The $108 billion guide is significant not because of what it says about NVIDIA but because of what it says about the companies writing the checks. Amazon, Microsoft, Google, and Meta have all signaled that their 2026 AI capex commitments are non-negotiable — and every dollar of that commitment eventually lands in NVIDIA's revenue line. When a single supplier can guide to $108 billion in one quarter with a ±2% confidence interval, it means its customers have already committed the purchase orders. This is not demand forecasting. This is order-book visibility.
Data center revenue at $89 billion in Q2 represented roughly 93% of total company revenue — a concentration that would concern analysts in any other sector but here functions as confirmation that the AI buildout is not broadening away from compute. It is deepening. The hyperscalers are not diversifying their spend; they are accelerating it through the same chokepoint. That dynamic is what makes NVIDIA's guide the single most important forward indicator in tech right now, more informative than any individual company's earnings call or capex disclosure.
The Broadcom Variable
The one live challenge to NVIDIA's monopoly narrative is Broadcom's custom ASIC business, and traders need to watch this week's fiscal Q3 report with that specific lens. Broadcom's deal with OpenAI — 1.3 gigawatts of custom processors pegged at $18 billion, contingent on Microsoft agreeing to purchase 40% of the chips — is not just a Broadcom story. It is a direct test of whether any hyperscaler or frontier AI lab is willing to commit capital to non-NVIDIA silicon at scale. Broadcom stock dropped more than 3% on the day of the report, a move that reflects genuine uncertainty about whether Microsoft will commit to the terms.
If Microsoft signs, the deal validates the custom ASIC thesis and introduces the first credible volume alternative to NVIDIA GPUs in the data center. If Microsoft walks, OpenAI must find replacement buyers for $7.2 billion worth of chips — a number that, if it evaporates, confirms NVIDIA's dominance for another hardware cycle. Broadcom's management commentary on ASIC pipeline, customer concentration, and co-investment terms will be more useful than any headline revenue figure the company reports. Traders holding AVGO or using it as a hedge against NVDA concentration risk should treat this report as binary.
What the Supply Chain Is Actually Telling You
The NVIDIA guide does not exist in isolation — it sits inside a supply chain that is simultaneously the most capital-intensive and most geopolitically exposed in the world. TSMC, which controls 72% of global foundry capacity and manufactures every major AI chip including NVIDIA's GPUs and Broadcom's custom ASICs, has already acknowledged that its U.S. capacity buildout is "far from enough." CEO C.C. Wei made that statement directly, and it carries operational weight: even with TSMC's $265 billion Arizona investment, advanced-node supply remains the binding constraint on how fast the $108 billion quarterly run rate can actually be fulfilled.
Layered on top of the supply constraint is the tariff variable. Commerce Secretary Howard Lutnick confirmed he is considering linking semiconductor tariff rates to the scale of each company's U.S. manufacturing investment — a framework that would create an explicit financial incentive to build domestically but that, in the near term, introduces cost uncertainty across every node of the AI hardware stack. If tariffs extend to products containing semiconductors, as currently proposed, the reach covers data center servers, which means the hyperscalers writing checks to NVIDIA could simultaneously face higher costs on the infrastructure side. That's a margin squeeze that does not show up in NVIDIA's revenue line but will show up in its customers' capex sustainability calculus by Q1 2027.
Anthropic's announcement that it has locked in $35 billion of NVIDIA-backed compute in Texas adds another dimension. That is not a purchase order in the traditional sense — it is a supply reservation, a bet that securing silicon access now is worth paying a structural premium versus waiting for market pricing. When a frontier AI lab spends $35 billion to lock supply before it needs it, the implicit message is that the people closest to the demand curve believe NVIDIA's allocation is going to get tighter, not looser, over the next 18 months.
The Micron earnings report on September 30 is the next hard data point. High-bandwidth memory pricing is the component-level read on whether HBM demand — which is essentially a derived demand from GPU unit volume — is holding at current levels or beginning to show the early signs of a digestion cycle. NVDA at $227.98 with a $108 billion guide baked in leaves very little room for a guidance miss or a tariff escalation before the multiple gets stress-tested. Traders long NVDA into Micron should define their exit level now: a close below $210 on tariff news or a Micron guide-down would be the first technical signal that the guide-driven momentum is breaking.
The Weekly Investor
Daily market analysis for active traders. Free.
