Broadcom Has Reached Memory Price Parity With NVIDIA
The firm secured volume-based pricing as it bankrolls massive AI infrastructure builds for startup Anthropic.
Updated on Oct. 1, 2026 in Semiconductors
Live Poll
Do you trust that large tech companies partnering on AI infrastructure will benefit consumers?
Broadcom has matched NVIDIA in memory chip procurement pricing following a surge in its purchasing volume. The company has also agreed to provide up to $42 billion in financing to Anthropic for the startup's AI infrastructure development.
Why it matters
This shift in procurement power signals increased competition for memory suppliers like Samsung, which is managing high pricing demands against NVIDIA's aggressive margin targets. The move supports Anthropic's multi-year scale-up, effectively banking the infrastructure required for the startup's massive upcoming compute deployments.
Broadcom achieved price parity due to high-volume procurement as Samsung faces pressure to sustain high memory prices to offset foundry operating losses. While NVIDIA currently reports 75% gross margins, Samsung expects its own operating losses to decline by approximately 42% in 2026.
The players
Broadcom
A global technology company that designs and supplies a broad range of semiconductor and infrastructure software products.
NVIDIA
A leader in AI computing hardware known for its high-performance GPUs and dominant market position in data center acceleration.
Anthropic
An AI research and development company focused on scaling large-scale compute infrastructure for generative model training.
Samsung
A South Korea-based conglomerate and major producer of semiconductor memory and foundry services.
The details
Broadcom leverages increased volume to negotiate bulk memory pricing directly with manufacturers like Samsung. This strategy counters the margin requirements set by competitors like NVIDIA, who have recently guided for slight margin compression from 75% down to a 71% to 72% range. Simultaneously, the $42 billion loan helps Anthropic satisfy its $125 billion in five-year lease commitments for AI hardware, fueling a massive build-out of TPU (Tensor Processing Unit — a specialized integrated circuit designed for machine learning acceleration) clusters.
Timeline
2026: Samsung operating losses are projected to drop by 42 percent.
2027: Anthropic is expected to become Broadcom's largest customer.
2027: Anthropic plans to deploy 5 gigawatts of TPUs.
2028: Anthropic plans to deploy 10 gigawatts of TPUs.
The Tech Race
This financing marks an acceleration in the race to secure massive, dedicated AI compute power at the gigawatt scale. Anthropic's planned deployment of 15 gigawatts of TPUs over the next four years positions the startup to rival the largest existing infrastructure footprints currently operated by major cloud providers.
This deal secures the long-term hardware pipeline for Anthropic's AI model development, potentially lowering compute overheads in future model releases. It indicates a massive commitment to TPU-based architectures, signaling which hardware platforms are prioritized for large-scale AI research and production.
The takeaway
Broadcom's parity with NVIDIA highlights the intense pressure on memory manufacturers to stabilize margins while supporting massive AI build-outs. Watch for reported operating loss figures in 2026 to see if Samsung's supply chain strategy successfully offsets its foundry business costs.
Further reading
For more on the industry's compute hardware trends, visit Semiconductors.
Source note: This article includes information reported by Wccftech.
Live Poll
Do you trust that large tech companies partnering on AI infrastructure will benefit consumers?
