Nvidia hits its first record since May, within $300B of $6T

Nvidia reached a record valuation for the first time since May, leaving it less than $300B away from becoming the first company valued at $6T. The rally was driven by demand for AI agents and a record $150B increase in its share-buyback program; ASML is also roughly $300B away from reaching a $1T valuation.

Analysis

NVDA’s marginal buyer is increasingly being underwritten by capital-return mechanics rather than a fresh revision to AI infrastructure demand. A buyback can tighten float and cushion drawdowns, but it does not solve the key valuation question: whether agentic-AI workloads translate into incremental GPU clusters rather than higher utilization of already-installed capacity. The near-term setup favors NVDA momentum, yet a multiple reset is likely if hyperscaler capex guidance shifts from capacity expansion toward efficiency and inference cost reduction.

ASML is the cleaner second-order beneficiary if AI demand extends the semiconductor capital-expenditure cycle into 2027-28, but its earnings torque arrives later than NVDA’s because foundry customers must first commit to leading-edge node capacity. That lag creates an opportunity: confirmation of TSMC, Samsung, or Intel EUV order strength can re-rate ASML before reported system revenue catches up. Conversely, a strong NVDA tape without corresponding foundry wafer-capacity commitments would signal that AI demand is being met through mix, software optimization, and installed-base utilization—not new lithography intensity.

Consensus may be underestimating the circularity risk in AI spending: vendor financing, cloud credits, and customer buybacks can support demand optics while weakening ultimate end-user ROI. Over the next 1-3 months, watch hyperscaler capex commentary and TSMC leading-edge utilization; over 6-18 months, the decisive variable is whether AI-agent revenue becomes large enough to sustain compute purchases after the initial model-training buildout. The thesis is falsified for the bullish semiconductor complex by consecutive capex-guide cuts from MSFT/GOOGL/AMZN/META or evidence that advanced-node utilization remains below the level required to trigger EUV capacity additions.

Key Decisions for Investors

• Maintain a tactical long NVDA position for the next 1-3 months, but use a trailing stop or defined-risk call spread rather than unhedged delta: capital-return support can extend momentum, while valuation downside is asymmetric if the next hyperscaler capex cycle disappoints. Reassess immediately after the next major cloud-provider earnings cycle.
• Initiate a 6-12 month long ASML / short SOXX pair in modest size if TSMC reports sustained advanced-node utilization and raises 2027 capex expectations. ASML offers more direct exposure to a durable leading-edge capacity build; short SOXX hedges broad semiconductor beta and memory/analog cyclicality. Exit if foundry capex guidance is deferred or export-control restrictions broaden materially.
• Use an NVDA / ASML relative-value monitor: add ASML only if NVDA outperforms by another 10-15% without a corresponding upward revision in foundry capex or EUV shipment expectations. That divergence would create a better entry for the equipment catch-up trade rather than chasing NVDA’s buyback-driven move.
• Do not underwrite the announced repurchase increase as incremental EPS support until funding source, execution cadence, and net-share-count trajectory are independently verified. If repurchases are largely offset by stock-based compensation, treat the signal as sentiment support rather than a fundamental catalyst.