Business & Strategy Questions

ASIC Encroachment at Hyperscalers: How does NVIDIA think about the pace at which hyperscaler custom silicon programs realistically displace NVIDIA GPUs, and what are the leading indicators to watch?

Context: Google, Amazon, and Microsoft collectively represent roughly half of Data Center revenue and are all investing heavily in internal ASICs. OpenAI has publicly claimed performance advantages for its "Jalapeno" chip despite NVIDIA's multibillion-dollar investment in the company. Management's counter-argument — that NVIDIA's platform is fungible across training, post-training, and inference while ASICs are workload-specific — has held so far, but the threat is growing in scale and credibility.

Frontier Lab Dependency and Circular Financing: How does NVIDIA think about the risk that AI labs where it has extended balance-sheet support slow their compute consumption before becoming self-funding — and what would the practical impact be on NVIDIA's revenue and guarantee obligations?

Context: Management expects AI labs where NVIDIA leverages its balance sheet to represent roughly a quarter of FY28 revenue. NVIDIA has disclosed maximum gross guarantee exposure of $108.5B as of Q2 FY27, including up to $105B tied to the SB Energy/OpenAI Portsmouth campus. The "fungibility" defense — that compute can be redeployed to other customers — has never been tested through an actual pullback in AI lab spending.

NeoCloud Utilization Risk: What happens to NVIDIA's take-or-pay revenue guarantees to NeoCloud lenders if GPU rental rates soften or NeoCloud utilization falls short of underwriting assumptions?

Context: NeoCloud capacity is expected to reach 8 GW by year-end 2026, up from roughly 3 GW at end of 2025, funded partly through a structure where NVIDIA provides minimum revenue guarantees to lenders in exchange for a share of rental revenue above a floor. Management acknowledges this segment is growing off a smaller, harder-to-underwrite base of capital-constrained operators. This is a new, unproven exposure that has not been tested through a downturn in GPU rental rates.

Vera CPU Competitive Positioning: What is NVIDIA's realistic path to becoming a leading server CPU supplier, given that x86 incumbents have decades of ecosystem depth and hyperscalers have their own Arm-based CPU programs?

Context: Management has guided to approximately $20B in standalone Vera CPU revenue in FY27 and expects CPU revenue to more than double in FY28. The $200B total server CPU TAM is a market NVIDIA has not historically participated in. The competitive dynamics — against AMD, Intel, and hyperscaler-designed Arm CPUs — are materially different from the GPU market, where NVIDIA has an established CUDA moat.

Groq LPU Market Sizing: How large is the addressable market for the Groq 3 LPX, and which specific customer segments or workloads justify the premium cost per token relative to standard Vera Rubin throughput?

Context: Management has described the LPX as a niche product suited for high-interactivity, high-ASP token services where latency matters more than throughput or context capacity. Jensen noted that LPX's SRAM architecture limits context processing and total throughput. The sizing question matters because NVIDIA has made significant non-refundable licensing payments to Groq that it may not fully recover if the addressable market is smaller than expected.

China Ecosystem Risk: How does NVIDIA assess the risk that Huawei's Ascend platform, having gained domestic scale, begins winning AI infrastructure contracts in markets outside China?

Context: NVIDIA has explicitly stated that China market foreclosure "helped our competitors build larger developer and customer ecosystems to challenge us worldwide." China's antitrust authority has also issued a preliminary finding tied to Mellanox acquisition conditions that could create additional friction. The longer NVIDIA remains excluded, the more durable and potentially exportable Huawei's ecosystem could become.

Physical AI Revenue Ramp: Physical AI has contributed over $9B in revenue over the last twelve months. What are the two or three most important milestones — technical, commercial, or regulatory — that would cause this business to inflect meaningfully faster?

Context: Automotive, robotics, and industrial physical AI are framed as the next major growth wave after agentic AI. Management has cited robotaxi fleets scaling from thousands to millions of vehicles, and factory digital twins as key drivers, but the revenue ramp has been gradual relative to Data Center. Understanding the gating factors helps investors assess the timing and magnitude of this opportunity.

Annual Architecture Cadence Execution Risk: As NVIDIA moves to a one-year product cadence across six or more chips simultaneously, how does management think about the risk of a stumble that delays a generation and creates a demand air pocket?

Context: NVIDIA itself acknowledges that the annual cadence "may magnify the challenges associated with managing supply and demand." Every major transition has produced at least one quarter of material margin compression. Rubin is the third-generation NVLink rack-scale system, but it introduces six new chips simultaneously, raising execution complexity significantly compared to earlier transitions.

Financial Questions

FY28 Revenue Bridge: Within the approximately 70% FY28 revenue growth guide, what are the rough relative contributions from Vera Rubin GPU volume growth, standalone Vera CPU, Groq LPX, and networking — and how much of the guide assumes pricing increases versus unit growth?

Context: Management has guided to approximately 70% FY28 revenue growth while describing unconstrained demand as closer to 100%. Understanding the composition of that guide — GPU versus CPU versus networking versus pricing — is essential for stress-testing assumptions and identifying where supply constraints are most binding. The Vera CPU alone is expected to more than double in FY28.

Gross Margin Path to FY28: Given that memory pricing increases "exceeded prior expectations" and margins are now guided to bottom at 71-72% in Q4 FY27 before settling at 72-73% in FY28, what specific mechanisms — pricing actions, mix shift, or cost structure improvements — give management confidence in the FY28 floor?

Context: The gross margin target has been reset downward repeatedly: from "mid-70s" to the current 72-73% FY28 guide. Management has cited executed price increases taking effect in Q1 FY28 as a key lever, but the magnitude of memory cost inflation has already surprised once. Investors need to understand whether the 72-73% FY28 guide is a floor or a midpoint, and what the sensitivity is to further memory cost increases.

Memory Cost Dynamics: How much of the memory cost inflation is driven by NVIDIA's own AI-driven demand surge versus broader industry supply constraints, and does NVIDIA have contractual protections or long-term agreements that limit further upside surprises?

Context: Management acknowledged that memory scarcity is "being driven in large part by the AI build-out itself," creating a situation where NVIDIA's own growth is inflating its input costs. NVIDIA has stated it has long-standing relationships with all three major memory suppliers, but the degree to which those relationships provide pricing predictability versus market-rate exposure is unclear.

Purchase Commitment Exposure: Total future purchase commitments reached $366B as of Q2 FY27, with $279B for supply and capacity. What is the expected timing of cash outflows against these commitments, and what portion is cancellable or renegotiable if demand undershoots?

Context: Inventory nearly doubled sequentially to $32B in Q2 FY27 ahead of the Vera Rubin launch. The scale of forward commitments is a key risk if end demand disappoints — the same dynamic that preceded the $4.5B H20 charge in Q1 FY26. Understanding the structure and flexibility of these commitments is critical to assessing downside scenarios.

Ecosystem Investment Cash Flow Impact: What is the expected cash outflow from ecosystem investments and guarantee commitments in FY28, and how does management think about the return threshold for these investments relative to share repurchases?

Context: Ecosystem investments totaled $17.5B in FY26 alone, and NVIDIA has committed to significant non-refundable licensing payments (e.g., Groq) and land/power/shell guarantees ($3.5B in FY26). The board has authorized $80B+ in incremental buybacks, but management has indicated it will prioritize strategic uses of capital. Understanding the expected FY28 cash consumption from ecosystem commitments versus buybacks helps investors model free cash flow conversion.

OpEx Trajectory: With full-year FY27 OpEx now guided to grow in the "low 50s" year-over-year, what is the right framework for thinking about OpEx growth in FY28 as the product portfolio broadens further?

Context: R&D grew 43% in FY26 to $18.5B, driven by employee growth, compute infrastructure costs, and engineering development for new architectures. The Vera Rubin, Rubin Ultra, and subsequent roadmap generations all require simultaneous engineering investment. Management has cited AI tooling as an efficiency offset, but the net trajectory of OpEx as a percentage of revenue is a key input to operating margin modeling.