Roberto Chamorro gilaberte
We’re moving toward a world where no one wants to depend on a single provider at any layer. And this applies across all three layers at the same time: Multicloud A few years ago, companies would choose one cloud (AWS, Azure, or Google) and put everything there. Now large enterprises distribute workloads across two or three clouds. Why? Negotiating power (if AWS raises prices, you can shift workloads to Azure), resilience (if one cloud goes down, your entire business doesn’t), and because each cloud excels at different things (AWS in infrastructure, Azure in Microsoft integration, Google in data and analytics). That’s why Uber runs across AWS, Google Cloud, and Oracle at the same time. And that’s why all three hyperscalers keep growing — it’s not a winner-takes-all market. Multichip Same concept, but in hardware. Until recently, it was Nvidia or nothing. Now you have Amazon’s Trainium, Google’s TPUs, Microsoft’s own chips (Maia), AMD gaining share, and Broadcom designing custom ASICs for companies like OpenAI. Companies want to diversify to avoid relying solely on Nvidia, negotiate better pricing, and optimize costs depending on the workload (training a model requires different chips than running inference at scale). Nvidia is still the clear leader, but it goes from ~95% share to maybe 60–70% over time. Still dominant, but the total market grows so much that it sells more chips than ever, even with lower share. Multi-model AI This is the newest layer. A year ago, many companies went all-in on a single model (GPT-4 from OpenAI, or Claude from Anthropic). Now the trend is to use multiple models depending on the task. A large and expensive model for complex reasoning, a smaller and cheaper one for simple tasks, one specialized for coding, another for document analysis. The loser in this world is the one trying to be a single, closed provider. The winner is the one offering the best platform where multiple options can coexist. Roberto Chamorro. $MSFT (Microsoft) $AMZN (Amazon.com Inc) $GOOGL (Alphabet Inc Class A) $META (Meta Platforms Inc) $NVDA (NVIDIA Corporation)
Not investment advice. The author may have financial interests in the mentioned instruments.
null
.