The dual-source becomes table stakes — Meta's 6GW AMD commit is the reference posture
Meta committing 6GW of AMD MI400 capacity in the same week as a parallel NVIDIA expansion makes dual-sourcing the structural hyperscaler default. NVIDIA's monopoly-pricing era is closing in measurable ways.
The number that matters
Meta's 6GW AMD commitment is roughly a year's worth of OpenAI's training compute, bought from the second source. The buyer just told the market that ~50% AMD share at training scale is the new procurement floor.
Why this is structural, not transactional
One-off deals come and go. Hyperscaler procurement decks getting line-itemed for second-source GPU capacity is the architectural shift. Three years ago, "NVIDIA or nothing" was the procurement default. Today, the procurement decks read:
- ~50% NVIDIA Vera Rubin (shipping Q4 2026)
- ~30–40% AMD MI400/Helios (shipping Q3 2026)
- ~5–10% Cerebras WSE / Groq LPU / custom ASICs for specialized workloads
Once those allocations show up in three hyperscalers' decks, the pricing-leverage math changes irreversibly.
The LPU wildcard
NVIDIA's own LPU announcement at GTC 2026 is the under-discussed move. The GPU vendor entering the language-specialized chip category is a tell: NVIDIA's strategists see Cerebras, Groq, and the custom-ASIC tier as a credible enough threat that they want their own offering in that lane. That's a defensive move, not an offensive one — and it acknowledges that the "general-purpose GPU" positioning isn't enough anymore.
NVIDIA shipping an LPU is the strongest signal that NVIDIA itself believes the substrate-diversification trend is durable.
The Q4 NVIDIA earnings test
The first cycle that fully reflects dual-source procurement dynamics is the Q4 2026 NVIDIA earnings call. Watch the data-center segment gross-margin line. If it holds at the current premium, the moat is intact. If it compresses by 2–4 points, the dual-source thesis is being priced in. The market will then re-rate NVIDIA against the AMD/Cerebras tier in a way the current multiple does not anticipate.
CNBC — Meta AMD 6GW → · DCD — NVIDIA LPU roadmap → · Futurum — Rubin + Helios CES →