Q&A with AMD executive Forrest Norrod on increasing AI investments, releasing accelerator chips at a faster cadence, AMD's open standards approach, and more
Dylan Martin / CRN : X: @dylanonchips , @techepiphanyyt , @dylanonchips , and @dylanonchips X: @dylanonchips : I recently talked to top @AMD data center exec Forrest Norrod about the company's decision to move to an annual data center GPU release cadence to fight Nvidia's AI dominance: “We see the opportunity is so great that we felt we had no choice.” Full Q&A: https://www.crn.com/... @techepiphanyyt : “And then [MI]350 ... we think is higher performance than what we see projected for B200. We think B200 is really a 2025 part for any sort of volume, and so is [MI]350” - Forrest Norrod (AMD) https://www.crn.com/... @dylanonchips : The Instinct MI350, on the other hand, is due out next year and is expected to have “higher performance than what we see projected for [Nvidia's] B200,” which Norrod considers a “2025 part for any sort of volume.” https://www.crn.com/... @dylanonchips : Forrest Norrod told me that @AMD's Instinct MI325X, set to launch in Q4, “handily outdoes [Nvidia's] H200 and is competitive in many regards with [the upcoming] B100.” https://www.crn.com/...
Context & Ripple Effects
AMD had already introduced the MI300 accelerator family with performance claims aimed at Nvidia’s H100-class systems. This interview turns that product push into a stated operating model: a yearly Instinct data-center GPU cadence backed by greater AI investment.
The significance is not just a single chip comparison. AMD is positioning product timing and open standards as complements to its hardware challenge, while Nvidia’s software ecosystem remains a key competitive constraint.
First-order effects
- AMD commits customers and channel partners to a faster, more predictable refresh rhythm for Instinct accelerators, beginning with the MI325X and followed by MI350.
- The company raises the competitive stakes around upcoming Nvidia systems by publicly framing its planned products as performance alternatives, while emphasizing an open-standards approach.
Second-order effects
- Enterprise buyers and cloud providers gain a clearer basis for qualifying AMD alongside Nvidia rather than treating Instinct as an occasional alternative; that can increase pressure on Nvidia’s upgrade timing and platform pricing.
- A yearly hardware cadence makes software readiness more consequential: AMD’s open approach must translate into supported tools and deployments quickly enough for each generation to be viable.
Third-order effects
- If AMD sustains the cadence, AI infrastructure buying could become less centered on a single accelerator roadmap and more on multi-vendor qualification, especially for workloads able to use heterogeneous compute.
- The contest increasingly shifts from peak-chip claims to an integrated platform race—hardware availability, software compatibility, and recurring upgrade execution—where Nvidia’s established CUDA position remains a structural hurdle for challengers.
The trend: AI accelerator competition is moving toward annual roadmap execution and full-stack ecosystem delivery rather than isolated benchmark-led product launches.