Date: 2026-03-05 Event: Morgan Stanley Technology, Media & Telecom Conference (March 4, 2026) Context: Jensen Huang fireside chat — follow-up to Q4 FY26 earnings (Feb 25)
Jensen did his fireside chat at the Morgan Stanley TMT Conference, and the headline everyone grabbed was the OpenAI/Anthropic investment pullback. The $30B OpenAI investment "might be the last." The $10B Anthropic stake "probably the last as well." The rumored $100B OpenAI mega-deal? "Probably not in the cards" — because both companies are heading to IPO.
The media is overplaying this. NVIDIA doesn't need equity stakes in these companies. Jensen said it: "If we provide the compute capacity they need, which we're ramping up hard... the revenues will more than follow." The investment was a strategic handshake. The real relationship is commercial — GPU sales, cloud compute, infrastructure deals. That hasn't changed.
1. Sovereign AI hit $30B+ in FY26 — tripled YoY. Countries building their own AI infrastructure: Canada, France, Netherlands, Singapore, UK leading. This is a durable demand layer independent of hyperscaler CapEx cycles. Diversification I hadn't fully appreciated in the Q3 review.
2. Agentic AI framed as the current inflection. Training → inference → agentic → physical AI. Agentic is more compute-intensive than simple inference — every reasoning step is another forward pass. More tokens = more GPUs. Good for NVIDIA's compute narrative.
3. Physical AI at $6B+ in FY26. Robotaxis (Waymo, Tesla, Uber, WeRide, Zoox), manufacturing, robotics. Small relative to $300B+ Data Center run rate, but represents entirely new TAM.
4. Morgan Stanley reinstated NVDA as top semi pick. Joseph Moore's call around the conference. After Q4 FY26 — $68.1B revenue (beat $66.2B consensus), +73% YoY, Data Center $62.3B — hard to argue against execution.
5. Q4 FY26 context (reported Feb 25). Record $68.1B quarter. Data Center $62.3B (+75% YoY). Continued beat-and-raise pattern. Vera Rubin samples already shipped to early customers, full production H2 2026.
| Question | Prior View (Q3 Review, Feb) | Updated View (Post-Conference) |
|---|---|---|
| Hyperscaler CapEx sustainability | Watching closely | Q4 delivered. $600B aggregate 2026 CapEx intact. Sovereign AI adds diversification. |
| ASIC risk timeline | Real for 2027+ | Unchanged. ~10% of AI processor revenue per Moore. NVIDIA ~85%. |
| OpenAI/Anthropic relationship | Deepening via investment | Shifting to pure commercial (GPU sales, not equity). Fine — revenue > equity. |
| FY27 margin picture | "Working to hold" cautious | Need GTC (March 16) for Vera Rubin economics |
| Demand durability | Strong but concentrated | Stronger — sovereign AI layer adds non-hyperscaler demand |
Thesis from Q3 review — tryout candidate at 4-5% — unchanged. If anything, the Q4 beat and sovereign AI diversification strengthen it slightly. But GTC on March 16 is the real catalyst:
The conference was maintenance, not a catalyst. Jensen said the right things. The investment pullback is a natural consequence of IPO timelines, not a strategic retreat.
Event: Neutral-to-slightly-positive. No new negative signals. Thesis: Intact. Tryout candidate. Sovereign AI adds conviction on demand durability. Next decision point: GTC (March 16-19). Vera Rubin economics and FY28 outlook.
As usual, thanks for reading.
Sources: CNBC, TechCrunch, PYMNTS, Investing.com, TheStreet, Motley Fool — March 4-5, 2026