AMD Denies Instinct MI455X Delay, Says H2 2026 AI GPU Shipments Remain on Track

AMD has pushed back against claims that its next-generation AI accelerator faces delays. The company insists the Instinct MI455X remains on schedule for 2026 despite reports suggesting manufacturing setbacks.

According to TechPowerUp, a SemiAnalysis report predicted a delay in AMD’s Instinct MI455X AI mass manufacturing. The forecast argued that AMD’s next-gen AI hardware could enter the market in a weaker position against NVIDIA’s upcoming “Vera Rubin” server GPUs.

AMD Calls Delay Report Inaccurate

AMD’s Corporate VP of AI Software and Solutions, Anush Elangovan, publicly rejected the claims. He described the postponement report as inaccurate and stated the project remains on track.

Elangovan confirmed that MI455X shipments are still planned for the second half of 2026. He also stressed that the final production model will not slip into 2027, directly countering rumors of engineering sample issues and manufacturing hiccups.

The MI455X forms a core part of AMD’s rack-scale “Helios” UALoE72 system, designed to compete in large-scale AI training and inference deployments. Elangovan highlighted rapid execution and platform bring-up speed as key strengths for the complex GPU rollout.

Competitive Pressure in the AI Accelerator Market

The debate underscores the intense rivalry between AMD and NVIDIA in the AI accelerator market. NVIDIA’s next-generation Rubin platform continues to dominate discussions around large-scale AI infrastructure, particularly as hyperscalers prepare for massive AI cluster expansions.

While AMD maintains that its timeline remains intact, questions linger about whether it can match NVIDIA’s large-scale factory output. Volume manufacturing and supply chain execution often determine real-world competitiveness as much as architectural design.

Beyond the MI455X discussion, AMD recently announced support for Alibaba’s Qwen 3.5 models on its Instinct accelerators, strengthening its AI software ecosystem.

At the same time, NVIDIA’s Rubin platform is driving demand for next-generation HBM4 memory. Samsung has confirmed its HBM4 plans and partnership with NVIDIA, signaling how memory vendors align closely with GPU roadmaps in the AI arms race.

As AI infrastructure spending accelerates globally, execution speed and manufacturing scale may prove just as decisive as raw performance numbers in determining who leads the next wave of data center AI hardware.