NVIDIA's stock price has remained nearly stagnant over the past two quarters, but Morgan Stanley's assessment is that the fundamentals have actually strengthened. The market is primarily held back by two concerns: whether growth can persist beyond 2026, and whether market share will gradually be eroded by ASIC and AMD. Morgan Stanley has reinstated NVIDIA as the top pick in the semiconductor sector, maintaining an Overweight rating and a $260 price target, viewing the current valuation as a rare entry window.
According to Morgan Stanley's North America semiconductor analyst Joseph Moore in a recent report, NVIDIA trades at only about 18 times 2027 earnings, representing a surprisingly attractive entry point. His logic is not based on betting on another earnings beat next quarter, but rather on the expectation that investor skepticism regarding "growth durability" will begin to ease in the coming months.
Evidence supporting this view comes more from supply chain and customer behavior: the report mentions that hyperscale cloud providers are locking in longer-term volume commitments with suppliers, with instances of "three-year orders, partially prepaid in full." Such actions are difficult to reconcile with plans for a sharp slowdown next year. However, the report also acknowledges that downstream cash flow pressure is a real issue: some cloud infrastructure businesses are in a negative cash flow state, and based on financial projections, capital expenditure could indeed slow next year.
Regarding market share, Morgan Stanley does not deny that NVIDIA might see a slight decline around 2026. Its breakdown suggests NVIDIA holds about 85% revenue share, ASIC slightly above 10%, and AMD slightly below 5%. Given NVIDIA's scale, now approaching "around $80 billion per quarter," it is not surprising that competitors might grow slightly faster, leading to a 1-2 percentage point share loss.
The essence of the two-quarter consolidation: the market is betting on "lack of durability," not denying current strength. Morgan Stanley attributes NVIDIA's recent price stagnation to expectations: the market already assumes current strength, with the real divergence of opinion concerning 2027. The report highlights an anomalous contrast – after switching the sector top pick from NVIDIA to SanDisk, and then to Micron, memory stocks have gained 300%-900% cumulatively, while NVIDIA's share price has been "stagnant." Meanwhile, earnings expectations for NVIDIA's "current quarter" have been raised by 38% over six months. This implies that the stock price is being held down not by data, but by doubts about the cycle's longevity. The report also places this sentiment shift in a historical pattern: over the past three years, the market has expressed concern about "the following year" early each year, until visibility improves, leading to a period of stock outperformance. It bets that 2026 will repeat this process.
Supply chain signals lean "hard": three-year lock-ins, even full prepayment. Morgan Stanley acknowledges that durability is harder to prove, but observes increasingly extreme upstream behavior. The report states that hyperscale customers are placing three-year orders with memory suppliers, in some cases even prepaying fully; an extreme case involves "receiving 100% of 2028 revenue this quarter," with the corresponding volume still being multiples of current levels. It views such prepayments as "durability clues": if customers truly planned to slow down next year, it would be difficult to explain why they would lock in far-future supply with cash upfront. The report emphasizes this is just one of many signs – supply chain feedback indicates multiple players are preparing for sustained growth.
The counter-argument from downstream is cash flow: some cloud infrastructure businesses are still "burning cash." Morgan Stanley does not downplay the "cash flow problem." The report notes that some cloud infrastructure businesses are in negative cash flow territory: this can certainly continue (strong balance sheets, financing options available), but based solely on financial numbers, next year's spending indeed "appears set to decelerate." The explanatory framework provided leans more towards a "timing mismatch": NVIDIA management's response on earnings calls is that cloud providers will generate more profit from these investments than market models suggest. The report adds that cloud GPU business profitability is high, and market participants discuss the "compute supply-demand gap widening by single-digit percentages daily." In other words, short-term cash flow might look poor, but that does not mean investment returns are absent.
Market share might see a slight dip, but customer focus is shifting to Rubin. On the competitive landscape, Morgan Stanley provides a clear breakdown: NVIDIA holds about 85% revenue share, ASIC slightly above 10%, AMD slightly below 5%. It expects competitors to grow slightly faster in 2026, with NVIDIA's share potentially declining by 1-2 percentage points, for a simple "physical" reason: larger scale brings stronger supply constraints, making it harder to maintain the same growth rate. But the report argues that customers are really focused on Rubin (expected to ship in the second half of the year), and until then, competition will primarily involve Blackwell versus various alternatives. It also acknowledges that the moat has "eroded somewhat," as leading model developers prefer architecture neutrality, using NVIDIA, AMD, Google Cloud TPUs, and custom ASICs concurrently. However, based on its research, even in areas where third parties emphasize "lower Total Cost of Ownership," customer preference often still leans towards NVIDIA. The report even mentions that the two largest ASIC users and the two largest potential AMD users might still see their business with NVIDIA grow by 80%+ in 2026.
Furthermore, regarding more aggressive bullish narratives (pointing to frontier model developers needing "hundreds of GW of compute" by 2029, compared to about 30GW this year), Morgan Stanley remains measured: it does not buy into this specific number, but clearly states it sees no signs of the cycle ending in 2026. The key constraint is capacity: semiconductor manufacturing capacity cannot ramp supply to that scale in a short time. More importantly, feedback from the global supply chain is not about "weakening orders," but rather "using cash to lock in growth through 2028 in advance."
For the upcoming GTC conference, Morgan Stanley expects it might not fully convince skeptical investors of the sustainability of capital expenditure, but will likely clarify the market share debate first. This showcase will be similar to 2024 – presenting a more complete four-year roadmap and emphasizing that competition involves not just the chip itself, but also rack-level solutions and ecosystem development. Simultaneously, it expects Groq IP to have a place in the roadmap.
Comments