AMD Has Begun Sampling MI450 GPUs & Also Engaged With Customers On MI500, Largest AI Deployments Are For Inference

AMD's MI450 GPUs are now being sampled to customers, and there's already huge interest in the MI500 series within AI Inference data centers.

Customers Are Lining Up To Deploy AMD's MI450 GPUs As Sampling Begins, Next-Gen MI500 Also Gains Traction In The "AI Inferencing" Segment

During the Q1 2026 earnings call, AMD CEO Dr. Lisa Su confirmed that they have started sampling the first Instinct MI450 GPUs to "lead" customers, and the chips remain on track for deployments at various data centers in the Helios rack, which will see production shipments commence in the second half of this year.

Several companies have already confirmed the deployment of AMD's MI450 series GPUs, with OpenAI and Meta signing multi-Gigawatt agreements. Anthropic is also reported to be in line to utilize the MI400 series to power its AI compute infrastructure.

We have begun sampling MI450 GPUs to lead customers and remain on track to ramp Helios production shipments in the second half of the year. As we approach production, demand for MI450 GPUs continues to strengthen, with lead customer forecasts now exceeding our initial plans, and a growing number of new customers engaging on large-scale deployment, including additional multi-gigawatt opportunities.

Dr. Lisa Su - AMD CEO

As for the customer interest surrounding MI450 AI GPUs and the Helios AI rack, there's lots of interest from major AI customers, a few names we mentioned above. AMD is already engaged deeply with the customers in what it calls "deep co-engineering".

The demand has already exceeded AMD's own internal expectations that they had planned for 2027, as such, MI450 is expected to see deployments at a significant scale across training and inference workloads, but the largest deployments will be made on the inference side as Agentic AI takes the helm of the new era of AI.

So, we are very excited about MI450 and Helios. We're seeing significant customer interest in those products as well. So, we have certainly talked about our large partnerships with OpenAI and Meta, and those are going really well. We appreciate the deep co-engineering that has gone on there.

You know, when we look at the totality of, let's call it, based on our current visibility, how those forecasts are coming in with all of our customers, we're actually seeing it above our initial plans that we had planned for 2027. And I think the encouraging thing is we're seeing a breadth of customers who are now very interested in deploying at significant scale MI450 series nodes. And those are for both training and inference workloads, although the largest deployments are for inference.

And, you know, based on all of that and the scale of new customer interest, you know, we see a path to really get to exceed our, you know, original, you know, targets of greater than 80% stagger. And these are really, you know, 2027 timeframe. Obviously, when we talk to customers, we're talking to them about, you know, MI355.

There's a lot of, you know, good traction we're seeing there. MI450 and Helios, I think, for significant large-scale deployments. And then many customers are also very engaged with us on the MI500 series and all of the opportunities there.

Dr. Lisa Su - AMD CEO

The AMD Instinct MI450 series lineup will be based on the CDNA 5 architecture and features the following highlights:

• Increased HBM4 Capacity & Bandwidth
• Expanded AI Formats with Higher Throughput
• Standard-Based Rack-Scale Networking (UALoE, UAL, UEC)

The official metrics list the MI400 as a 40 PFLOP (FP4) & 20 PFLOP (FP8) product, which doubles the compute capability of the MI350 series, which is a hot product for AI data centers.

In addition to the compute capability, AMD is also going to leverage HBM4 memory for its Instinct MI400 series. The new chip will offer a 50% memory capacity uplift from 288GB HBM3e to 432GB HBM4. The HBM4 standard will offer a massive 19.6 TB/s bandwidth, more than double that of the 8 TB/s for the MI350 series. The GPU will also feature a 300 GB/s scale-out bandwidth/per GPU, so some big things are coming in the next generation of Instinct.

AMD has positioned its Instinct MI450 GPUs against NVIDIA's Vera Rubin, and the high-level comparison looks something like the following:

• 1.5x Memory Capacity vs Competition
• Same Memory Bandwidth vs Competition
• Same FP4 / FP8 FLOPs vs Competition
• Same Scale-Up Bandwidth vs Competition
• 1.5x Scale-Out Bandwidth vs Competition

For the MI400 series, there will be two products, the first one is the Instinct MI455X, which is aimed at scale AI Training & Inference workloads. The other one is MI430X, which is aimed at HPC & Sovereign AI workloads, featuring hardware-based FP64 capabilities, hybrid compute (CPU+GPU), and the same HBM4 memory as the MI455X.

I look forward to sharing more on our next-generation Instinct GPUs, EPYC processors, Helios' RackScale platform, and our growing customer engagement at our Advancing AI event in July.

Dr. Lisa Su - AMD CEO

AMD will be offering more details on the Helios Rack at its upcoming "Advancing AI" event, which will be held in July.

AMD is gaining strong momentum in the AI accelerator market, with its MI450 GPUs entering customer sampling and poised for large-scale deployments in the second half of 2026. Backed by robust demand from major players like OpenAI, Meta, and most likely, Anthropic — exceeding initial 2027 forecasts — the MI450 series is seeing significant traction in both training and especially inference workloads.

Featuring major leaps in HBM4 memory, compute performance, and rack-scale capabilities, the platform positions AMD as a competitive alternative to NVIDIA, while early interest in the next-generation MI500 series further strengthens its long-term outlook in AI infrastructure.

Follow Wccftech on Google to get more of our news coverage in your feeds.