Billionaire investor Stanley Druckenmiller sold his Broadcom stake and bought Amazon and Alphabet. He sees hyperscalers gaining control over their own chip supply.
Amazon and Alphabet trade at low valuations despite heavy capital spending.
Their large revenue backlogs may support future returns.
What happened
Stanley Druckenmiller's Duquesne Family Office fully exited its Broadcom position last quarter and bought large stakes in Amazon and Alphabet instead.
Why it matters
Hyperscalers like Amazon and Alphabet are taking more control over which chips go into their data centers, favoring their own custom silicon—like Amazon's Trainium and Alphabet's TPUs—over off-the-shelf options. This shift may pressure chipmakers like Broadcom even as it partners with these firms.
What to watch
Amazon trades at 20.5 times forward earnings and Alphabet at 16.4 times, with both showing negative free cash flow due to massive data center spending. They hold contracted revenue backlogs of $496 billion and $514 billion, respectively, which could support strong returns once new capacity comes online.
Ask the AI about this article →
Druckenmiller's move reflects a broader shift in AI compute: the largest cloud providers are increasingly designing and deploying their own chips rather than relying solely on external suppliers. Amazon's Andy Jassy said most new chips in its data centers this year will be its Trainium, while Alphabet is even selling TPU systems externally. This gives hyperscalers more control over their supply chains and potentially better economics.
The article frames this as a long-term trend favoring those with control over chip design and deployment. While Broadcom makes networking and custom AI chips, including TPUs, customers like Amazon and Alphabet are diversifying sources—Alphabet recently added Marvell for inference TPUs. This could pressure Broadcom's growth even as overall AI demand rises.
For investors, the timing is notable: both Amazon and Alphabet trade at historically low forward P/E ratios (20.5 and 16.4 times), reflecting fears about negative free cash flow from heavy capital spending. Yet their massive contracted revenue backlogs—$496 billion and $514 billion—suggest strong future demand. Druckenmiller's bet appears to be that these companies will turn that spending into durable returns as data centers come online.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
OpenAI announced on August 28 that it will terminate its model access contract with AI coding tool Cursor, pro…
Bernstein analysts say AI is creating a memory bottleneck, with demand expanding from high-bandwidth memory (H…
OpenAI CEO Sam Altman said in a Time magazine interview that he thinks "it is a good time to slow down" AI mod…
Nvidia's earnings on Wednesday shifted the narrative, showing its advantage extends beyond GPUs to the systems…
Deepgram has introduced Enhanced Metrics for Amazon SageMaker AI, a service by Amazon Web Services (AWS)
Nvidia has validated Micron's biggest AI bull case, according to a Seeking Alpha report
