AMD · Instinct MI400 · Layer 2
AMD Helios (Instinct MI455X)
Buyers with a working ROCm story who want more memory per GPU than NVIDIA sells, and a second source at rack scale.
rent Rentable by the hour from multiple clouds; purchasable at scale.
Specifications
- Memory
- 31 TB of HBM4 per rack across 72 MI455X GPUs; 432 GB of HBM4 per GPU.
- Bandwidth
- 260 TB/s scale-up per rack; 19.6 TB/s per GPU.
- Compute
- 2.9 exaFLOPS FP4 and 1.4 exaFLOPS FP8 per rack, on AMD's own figures.
- Power
- Not published per rack.
- Interconnect
- Scale-up fabric, Helios rack architecture
- Software
- ROCm
- Form factor
- rack
- Workload
- both
Verified 2026-09-09. Fields reading “not published” are exactly that — we do not estimate a figure a vendor withholds.
The catch
ROCm is still the decision, not the silicon. The memory advantage per GPU is real and large — 432 GB against Blackwell-class parts — but it only pays if your stack runs on ROCm without a rewrite, and for most teams today it does not. Racks are quoted around $5-5.5M, which is one of very few public rack prices anywhere at this layer.
The economics
AMD's leverage here is memory capacity and being a second source, not price per FLOP. The commercial proof is commitment rather than benchmark: Oracle Cloud is the first hyperscaler offering a public MI450-series supercluster at 50,000 GPUs from Q3 2026, and OpenAI has committed to 6 GW of AMD Instinct with the first gigawatt in H2 2026.
Cost per million tokens is the metric that decides most real purchases, and there is no neutral benchmark for it — every published figure comes from a company selling one side of the comparison. Read all of them, ours included, as claims rather than measurements.
The power question underneath this
A rack of frontier accelerators draws 120–200 kW against a 2026 average of about 27 kW, and the US grid interconnection queue exceeds 2,600 GW with waits approaching five years. Whether you can energise this part is now a harder question than whether you can buy it.
Layer 1 — Energy · The interconnection queue · Rack power density · Tokens per watt
Where you actually rent this
Choosing the part is the easy half. The same accelerator rents for wildly different money depending only on who you rent it from — the H100 spread across published on-demand rates runs about 9x — and the provider, not the chip, is what decides whether you can leave later.
Compared against
- NVIDIA Vera Rubin VR200 NVL144 vs AMD Helios (Instinct MI455X) — The 2026 rack-scale fight. Both bet everything on HBM4, and HBM4 is sold out through the year.Compare with NVIDIA Vera Rubin VR200 NVL144
- AMD Helios (Instinct MI455X) vs NVIDIA GB200 NVL72 — 432 GB per GPU of ROCm against the CUDA rack everyone already runs. The memory argument at rack scale.Compare with NVIDIA GB200 NVL72
Related silicon
- AMD Instinct MI300X — 192 GB HBM3, rent
- AMD Instinct MI350X — 288 GB HBM3E, rent
- AMD Instinct MI355X — 288 GB HBM3E, rent
- NVIDIA B200 — 192 GB HBM3E, rent
- NVIDIA GB200 NVL72 — 13.5 TB HBM3E across the rack, rent
- NVIDIA GB300 NVL72 — Higher HBM3E capacity than GB200; per-rack figure not consistently published, rent