AMD Q2 2026 earnings: record revenue, stock falls on AI concentration fears

The dominant framing of AMD's sell-off was 'investors worried about AI customer concentration,' but the numbers point to a more specific culprit: capital spending. AMD's Q2 capital expenditure surged to $808 million, nearly tripling from $282 million a year earlier ^[1]- far outpacing the 50% year-over-year growth in total revenue and even the 107% year-over-year growth in Data Center revenue. That spike hit free cash flow directly. Yet AMD guided Q3 non-GAAP gross margin to roughly 56%, flat with Q2 despite continued sharp revenue growth ^[2]- a signal the market read as an absence of margin expansion even as the company scales. Put together, the capex jump and flat-margin guide gave the sell-side a concrete, mechanical reason to trim numbers that had little to do with whether OpenAI or Meta stay loyal customers; the 'concentration fear' narrative was the headline, but the capex line was the trigger.

AMD didn't miss - it beat revenue and EPS estimates and guided Q3 revenue to approximately $13 billion, plus or minus $300 million ^[3], above consensus expectations ^[4]. The problem was the setup: AMD shares had already climbed 21% in the five trading days heading into the print, pricing in an outcome bigger than what the company ultimately delivered ^[4]. Q3 guidance topped consensus but fell short of the market's most aggressive, optimistic estimates ^[4], which is enough to trigger a sell-off when a stock has run that hard beforehand. The reversal was violent: AMD closed the regular session up roughly 7% only to fall as much as 8.94% after hours, erasing close to $70 billion in market value in a matter of hours ^[5]. Bernstein's Stacy Rasgon captured the dynamic bluntly, noting investors 'may have been looking for more' after Intel's own strong report raised the bar for the whole sector just ahead of AMD's print ^[6]. Futurum Group's Daniel Newman went further, saying the market wanted a 'blowout guide driven by Helios' ^[6]- AMD delivered a strong guide, just not a blowout one. In a stock priced for a blowout, strong reads as a disappointment.

Part of what makes AMD's AI customer relationships different from a normal supply deal is how they're structured. OpenAI and Meta each hold performance-based warrants to buy up to 160 million AMD shares at one cent apiece ^[7]- near-free equity that vests as the customers hit deployment milestones, aligning the customers' incentives with AMD's stock price rather than just locking in a purchase-order relationship. Combined, that's up to 320 million warrant shares, which could push AMD's total share count toward roughly 1.98 billion and dilute existing shareholders by an estimated 16-19.6% if fully exercised ^[7]. AMD is also funding the other side of the relationship: the company has committed to invest as much as $5 billion in Anthropic tied to deployment milestones, with Anthropic's Helios rack-scale deployment covering up to 2 gigawatts of MI450 systems ^[8]. That is a meaningfully different economic arrangement than simply 'selling chips to OpenAI, Anthropic, and Meta' - AMD is trading equity upside and capital commitments for demand commitments, and at least one sell-side analyst has stayed Neutral specifically because of AMD's 'significant exposure to OpenAI and elevated operating expenses,' wanting clearer visibility into Meta and OpenAI deployment timing through 2027 ^[9]. The AI-customer-concentration risk investors cite isn't just about revenue mix; it's about how much of AMD's future share count and balance sheet is now tied to a handful of counterparties delivering on multi-year, milestone-gated promises.

What's notable about the post-earnings analyst reaction is that it converged closely on the fundamentals even where it split on the price. Morgan Stanley raised its 12-month price target to $465, citing strength in the CPU business and AMD's Helios positioning ^[1], and that optimism was shared by most of the sell side. But the agreement wasn't unanimous: at least one analyst held a Neutral rating specifically citing AMD's 'significant exposure to OpenAI and elevated operating expenses' ^[9], wanting clearer visibility into Meta and OpenAI deployment timing through 2027 - a genuine disagreement about customer concentration risk, not just a question of valuation timing. Scale reinforces that caution: even with Data Center revenue up 107% year-over-year to $6.7 billion, AMD remains far behind Nvidia's roughly $75.2 billion in data center revenue ^[10], a gap that keeps some investors skeptical of how much ground AMD has actually closed. CEO Lisa Su framed the results differently, calling Data Center revenue more than doubling year-over-year part of 'an excellent quarter' and pointing to AI-driven demand expansion across all of AMD's markets ^[3]. Outside that Neutral holdout, the remaining dissent wasn't about whether AMD's AI business is real - it was about whether the stock's run-up already reflected it. That's consistent with a negative competitive data point that surfaced right around the print: reporting that Elon Musk said SpaceX would commit fully to Nvidia's architecture over AMD for space data center plans added a competitive headline right when investors were already primed to sell strength ^[6]. The split isn't just Su versus the analysts - it's a market that largely agrees on the growth story but disagrees over concentration risk and how much of that growth was already priced in before the print.