Nvidia record fuels AI accounting scrutiny

Nvidia shares touched an all-time intraday high of $237.88 last week after Morgan Stanley reinstated the company as its top semiconductor pick with a $300 price target. The rally reflects continued confidence in AI computing demand, backed by Morgan Stanley’s view that Nvidia’s customer base extends well beyond hyperscalers to model developers, neocloud providers, sovereign buyers, ODMs, OEMs, and enterprises.

A central dispute is now the useful life of AI hardware. Nvidia argues that major customers such as Meta and Alphabet’s Google have extended AI server depreciation from roughly four years to a six-year schedule, while Michael Burry says older chips could lose value quickly as more efficient processors arrive. Nvidia has pointed to CUDA compatibility across generations as a reason existing systems may avoid becoming stranded assets.

Investor scrutiny is also shifting to hidden leverage behind the AI buildout. Needham analyst Laura Martin said Meta has roughly $628 billion in off-balance-sheet liabilities, which would lift its valuation multiple by 35% if folded into enterprise value. Applying a consistent version of that approach to Alphabet pushes Google’s adjusted EV/FY27E revenue multiple to 9.0x, above Meta’s 8.3x.

Hyperscaler commitments now total $3.8 trillion, raising questions about whether valuations fully reflect future payment obligations. Nvidia’s next catalysts include supplier earnings, results from Microsoft, Meta, Amazon, and Alphabet, and its upcoming GTC event, where CEO Jensen Huang is expected to unveil new products across GPUs, networking, and software.