
Broadcom is a steady, contract-backed AI infrastructure supplier.
AMD is a higher-risk bet on taking GPU market share from Nvidia.
Broadcom's backlog locks in future revenue; AMD offers transformative upside if it wins.
What happened
Broadcom's AI semiconductor revenue sits near $8.4 billion per quarter, with a disclosed AI chip backlog of around $73 billion. Management now sees line of sight to more than $100 billion in AI revenue in 2027.
Why it matters
Broadcom is treated as a contracted infrastructure utility, while AMD is priced like a call option on gaining GPU market share from Nvidia. AMD's forward P/E ratio is much higher because it has a small slice of the AI accelerator market now, but its data center revenue could jump from several billion to tens of billions if it wins real share.
What to watch
AMD's MI355X packs more memory than Nvidia's B200 and shows better throughput and lower cost per token on some big language model tests when workloads fit on fewer GPUs. But Nvidia's Rubin platforms, which began shipping this summer, cut token costs by as much as tenfold versus older Blackwell setups.
Ask the AI about this article →
The article lays out why the market assigns such different valuations to two chipmakers riding the same AI wave. Broadcom's business is anchored by multiyear contracts and custom ASIC designs for hyperscalers like Microsoft, Alphabet, Amazon, and Meta. That contracted revenue stream makes it look like a reliable infrastructure utility, which supports a solid but not wild valuation. In contrast, AMD is a challenger trying to displace Nvidia's dominance in data center GPUs, with its Instinct MI350 and MI355X accelerators and Helios racks. Its potential to capture meaningful share is large, but uncertain, which explains its higher-risk, higher-reward stock pricing.
A key comparison is in the system-level performance. While AMD's MI355X shows advantages on some tests, Nvidia's Rubin platforms, which began shipping this summer, cut token costs by as much as tenfold versus older Blackwell setups and leverage years of CUDA and TensorRT tooling. This means AMD must prove that its Helios racks and ROCm ecosystem can offer the same end-to-end reliability and developer comfort as Nvidia's full stack. The article's takeaway is that investors are paying for Broadcom's steady cash flows and AMD's speculative upside, not just their current financials.
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.
Panasonic Holdings announced on August 28, 2026, that it will reorganize its technology divisions effective Oc…

SanDisk and Kioxia plan to invest over $31 billion over six years in Japan to expand NAND flash production, in…

OpenAI's ChatGPT Work, available to $20/month and up subscribers, splits into a cloud product accessible via c…

Applied Materials CEO Gary Dickerson said AI is driving the semiconductor industry's most consequential growth…

Social media conversations about Tempus AI intensified after Moderna and Merck reported Phase 3 success in per…

Researchers introduced Agent Seer, a pipeline that automatically creates realistic test scenarios for AI agent…
