Everyone's Wondering What's Next for Nvidia. This News From Amazon and Alphabet Offers Us an Idea That's Strikingly Clear

Nvidia's annual sales rose to $215 billion from $60 billion two years ago, and the article argues demand for its AI platforms remains strong even as Amazon and Alphabet expand their own chips. Amazon said it still plans to order "substantial quantities" of Nvidia chips, while Alphabet will begin delivering TPUs to select customers and recognize revenue from that business next year. Overall, the piece suggests Nvidia's growth can continue despite rising in-house competition from major cloud providers.

Analysis

The key takeaway is not that Nvidia is losing share, but that the AI compute market is broadening faster than the single-vendor narrative implied. Cloud hyperscalers are effectively turning into both buyers and competitors: they still need Nvidia for frontier workloads, but their proprietary silicon is increasingly the economic choice for steady-state inference and price-sensitive enterprise usage. That creates a bifurcated market where Nvidia captures the highest-performance tier while Amazon and Alphabet monetize lower-cost internal alternatives and keep more margin in-house.

The second-order effect is on pricing power. If the largest cloud buyers can credibly substitute even 15-25% of incremental demand into internal chips over the next 12-24 months, Nvidia’s growth can remain strong while gross-margin expansion becomes harder to sustain. The more important question is not unit demand, but mix: the market may be underestimating how much of Nvidia’s future growth shifts from hyper-scarcity pricing to a more normal competitive cadence as custom silicon matures.

For Amazon and Alphabet, this is a margin defense story before it is a share-gain story. Every workload migrated to in-house accelerators reduces external COGS and improves cloud economics, which should matter more to valuation than headline revenue from selling their chips to third parties. The adjacent beneficiary is the semiconductor ecosystem around networking, packaging, and memory, where demand remains levered to total AI deployment even if GPU concentration eases.

The contrarian view is that consensus may be too focused on Nvidia losing monopolistic status and not focused enough on the size of the expanding pie. In the next 6-12 months, Nvidia can still outperform if the market keeps rewarding absolute growth and capex intensity. But over a 12-24 month horizon, the cleaner trade may be relative value: own the hyperscalers with self-supply leverage and fade any re-rating of Nvidia that assumes perpetual scarcity economics.

Key Decisions for Investors

• Stay long NVDA tactically over the next 1-3 quarters, but trim into strength on any multiple expansion tied to scarcity rhetoric; upside remains tied to absolute AI capex, while downside is a future mix shift.
• Initiate a 6-12 month long AMZN / short NVDA pair trade if cloud capex remains elevated but proprietary chip adoption continues; thesis is margin retention at AWS versus eventual normalization in Nvidia’s pricing power.
• Add to GOOGL on 12-24 month horizon as TPU commercialization improves cloud margin structure; risk/reward is better than headline AI suppliers because monetization accrues internally before it shows up externally.
• Use call spreads on NVDA rather than outright longs for the next earnings cycle; the stock can still grind higher, but implied upside is increasingly constrained by the market’s awareness of custom-silicon substitution.
• Watch for a pullback in semi-capex names tied to hyperscaler spending if AMZN/GOOGL signal faster in-house chip adoption; that would be the early warning that external GPU demand is normalizing faster than consensus expects.