DOJ Subpoenas NVIDIA; Google AI Remedy Expands
DOJ issues subpoenas in NVIDIA antitrust probe as Google's search remedy now covers generative AI firms. FTC advances Microsoft cloud investigation.
Key Points
- The DOJ has escalated its NVIDIA antitrust investigation to the subpoena stage, compelling document production from NVIDIA and unnamed third-party companies — a step that historically precedes formal enforcement by 12–24 months.
- The Trump DOJ simultaneously argued that Google's search antitrust remedy should cover generative AI firms, and the court accepted that extension, creating a new judicial precedent for AI market oversight.
- With the FTC also advancing a broad probe into Microsoft's cloud and AI businesses, three of the five most valuable U.S. tech companies now face concurrent, escalating federal antitrust actions.
The DOJ has issued subpoenas in its antitrust investigation into NVIDIA — not just to the company itself, but to third-party firms — a material escalation that moves the probe from information-gathering into compelled legal process. NVIDIA stock was trading at $228.52, up 0.97%, as the news crossed Wednesday, a market reaction that reflects either investor complacency about regulatory tail risk or a calculated bet that enforcement remains 12–24 months away. Neither interpretation makes the subpoena stage irrelevant to traders holding a stock at 63x forward earnings with only 8% consensus upside baked in for Intel — and 45% for NVIDIA itself.
NVIDIA: From Inquiry to Subpoena
The distinction between a regulatory inquiry and a subpoena is not semantic — it is procedural and legal. An inquiry allows companies to respond selectively and on their own timeline. A subpoena compels production, carries legal penalties for non-compliance, and signals that the DOJ has identified specific documents, communications, or business arrangements it believes are material to a potential case. The extension of subpoenas to third-party companies is the more significant detail: it suggests investigators are building a picture of how NVIDIA's market position affects customers, competitors, or partners in ways that NVIDIA's own records alone cannot establish.
The core antitrust theory around NVIDIA centers on its dominance in AI training chips, specifically the H-series and Blackwell GPU architectures that account for the overwhelming majority of hyperscaler AI infrastructure spend. NVIDIA's CUDA software ecosystem — which locks developers into NVIDIA hardware by making it computationally expensive to migrate workloads — is the structural feature most likely to attract sustained antitrust scrutiny. A formal DOJ complaint, if one eventually materializes, would almost certainly focus on whether NVIDIA used CUDA exclusivity to foreclose competition from AMD, Intel's Gaudi accelerators, and custom silicon from Google, Amazon, and Microsoft. The subpoena stage historically precedes formal enforcement action by 12 to 24 months, which puts potential case filing between Q4 2027 and Q4 2028 — well beyond the current earnings cycle but well within the horizon of any long-term NVIDIA position.
What makes the DOJ action more complex is the simultaneous emergence of AMD as a credible competitor. AMD recently secured a deal with Meta worth up to $60 billion and has AI compute agreements with both OpenAI and Anthropic. AMD's Q2 non-GAAP EPS surged 246% year-over-year to $1.66 on revenue of $11.5 billion, up 50%. If AMD continues to win hyperscaler contracts at this pace, it partially undermines the DOJ's market definition argument — a market with a credible number-two player and active customer switching is harder to characterize as monopolized. That dynamic cuts both ways: it may reduce DOJ urgency, or it may cause investigators to focus more narrowly on the CUDA software layer rather than chip hardware, where competition is more visible.
Google, Microsoft, and the Regulatory Stack
Wednesday's regulatory picture extends well beyond NVIDIA. The Trump DOJ argued in the Google search antitrust proceeding that the court's remedy should explicitly cover generative AI firms — framing GenAI as "the possibility of the next digital frontier" that Google's search dominance could extend into. The court accepted the DOJ's position, bringing qualified AI firms within the scope of judicial oversight for the first time. The practical implication is significant: any generative AI company that receives preferential access to Google's search data, distribution, or advertising infrastructure could now be subject to court-supervised conditions. That touches Gemini directly, but it also affects any AI startup that depends on Google Cloud credits, API integrations, or distribution through Google's app ecosystem.
Google's own AI expansion is accelerating in parallel with its legal exposure. The company launched Gemini 4 Argon this week to paid subscribers on AI Pro and AI Ultra plans, with API and AI Studio access also opening. Argon is positioned for "complex, multistep tasks" — Google's language for enterprise and agentic workflow use cases that compete directly with Microsoft Copilot and Anthropic's Claude for enterprise contracts. The irony of Google's position is acute: it is simultaneously the defendant in the most consequential tech antitrust case in a generation and the company rolling out its most capable AI model to date. The court's expansion of the remedy scope to GenAI means that Gemini's commercial terms, distribution agreements, and data access arrangements are now potentially subject to judicial review.
The FTC's advancing probe into Microsoft adds a third front to what is becoming a coordinated — if not formally coordinated — federal antitrust push against AI incumbents. The Microsoft probe, initiated under the Biden administration and now advancing under the Trump FTC, covers cloud, AI, and software businesses. Microsoft's Copilot integration across Office 365, Azure, and Teams creates bundling arrangements that are structurally similar to the browser bundling case that defined United States v. Microsoft in the late 1990s. The FTC's separate proposed policy statement on AI accuracy — arguing that state AI laws requiring alteration of AI model outputs are "impliedly preempted" by federal regulatory schemes — simultaneously gives large AI labs deregulatory cover at the state level while preserving federal oversight authority. That is not a contradiction; it is a consolidation of regulatory jurisdiction from fifty state frameworks into one federal one.
What the Regulatory Calendar Means for Positions
The EU is moving on a parallel but distinct track. Brussels is considering expanding the Digital Markets Act to classify AI businesses as "gatekeepers," which would mandate interoperability between AI systems — a requirement that cuts directly against the closed-ecosystem strategies of Google (Gemini), Meta (Muse), and Apple (Apple Intelligence). Interoperability mandates, if enacted, would commoditize AI interfaces and shift competitive advantage toward underlying model quality and compute infrastructure rather than distribution lock-in. That is a structural negative for platform companies and a structural positive for independent model developers and API-first AI businesses.
Intel's position in this regulatory environment is paradoxically interesting. Intel sits 20% below its September peak near the $100 support level, with Q3 earnings due October 29. Its GPU D270 — a 480GB card debuted for AI inference at OCP this week — represents Intel's most credible recent challenge to NVIDIA's inference incumbency. If DOJ action against NVIDIA eventually constrains the company's ability to enforce CUDA exclusivity, Intel's GPU business becomes a primary beneficiary. That scenario is 12–24 months away at minimum, but it is the kind of asymmetric regulatory optionality that deep-value traders in INTC are beginning to price. The $100 level is the line to hold into October 29 earnings; a break below it on volume before then would signal that the market is not waiting for regulatory catalysts.
For NVIDIA, the immediate trading question is whether $228.52 already discounts the subpoena risk. At 63x forward earnings with 45% consensus upside, the stock is priced for continued AI infrastructure dominance with minimal regulatory friction — an assumption that Wednesday's DOJ escalation directly challenges. The next formal milestone is unclear, since DOJ investigations at the subpoena stage do not follow predictable public timelines. But any Commerce Department guidance on AI chip export controls — which the Chinese market data point complicates further, given that Chinese firms bought over one million NVIDIA RTX Pro 6000D and 5000 chips in 2026 — could arrive before year-end and create a near-term catalyst that operates independently of the antitrust timeline. Watch for Commerce Dept. guidance and any third-party subpoena recipients becoming public, either of which would reprice NVDA regulatory risk in real time.
The Weekly Investor
Daily market analysis for active traders. Free.
