AMD introduced Helios on July 20, 2026, describing it as the company's first rack-scale (Rack-scale) AI computing system and its first complete AI infrastructure product capable of competing directly with NVIDIA's Grace Blackwell and the next-generation Vera Rubin systems, according to TechNews. The system is expected to begin shipping later this year (E1).
Helios integrates four components: AMD Instinct GPUs, EPYC CPUs, high-speed networking, and the ROCm software platform, allowing it to be deployed directly in data centers to support large-scale AI model training and inference, per TechNews (E2). Liberty Times separately reported that AMD's first AI rack system, Helios, will ship in the second half of this year (E8) — consistent with TechNews' "later this year" timeline.
Microsoft announced on the same day, July 20, that it has become the newest Helios customer, according to TechNews. The company will deploy Helios in its Azure data centers as a computing platform for frontier AI model inference, Azure AI services, and enterprise applications, and will adopt AMD's newest EPYC "Venice" processor to support agentic AI, data processing, and semiconductor design workloads (E4).
Microsoft CEO Satya Nadella stated, per Liberty Times: "We are expanding our Azure infrastructure portfolio with AMD's Helios, delivering the performance, scale and choice customers need to build and run next-generation AI applications" (E9). Liberty Times also reported that Microsoft will launch two new computing instances built on AMD's Venice CPU — one targeted at agentic AI and data-processing pipelines, and the other at semiconductor design (E10).
Beyond Microsoft, TechNews reported that Meta, OpenAI, Oracle, and India's Tata Consultancy Services (TCS) have all committed to deploying Helios (E5).
AMD CEO Lisa Su said, according to Liberty Times, that "when it comes to inference, memory bandwidth and memory capacity," Helios has "a significantly more compelling advantage" compared with NVIDIA's rack-scale systems (E12).
On pricing, AMD has declined to disclose Helios' price, but research firm Futurum Group estimated it at between $5 million and $5.5 million, compared with an estimated $3.5 million to $4 million for NVIDIA's second-generation rack system, Vera Rubin, per Liberty Times (E13).
| System | Estimated Price (Futurum Group) |
|---|---|
| AMD Helios | $5 million–$5.5 million |
| NVIDIA Vera Rubin (2nd-gen) | $3.5 million–$4 million |
Research firm Futurum estimated, as reported by TechNews, that NVIDIA currently holds more than 95% of the data center GPU market, while AMD holds around 4.5% — but as large cloud providers push to diversify their supply chains, Helios could become a key lever for AMD to expand its AI market share, with potential to reach 20% to 25% (E3). Liberty Times reported the same underlying figures from Futurum Group: NVIDIA's share above 95% versus AMD's 4.5%, adding that Helios "could change that picture" (E14).
AMD said that 8 of the top 10 AI companies run workloads on its Instinct GPUs, including OpenAI, Cohere, and Elon Musk's SpaceX AI, according to Liberty Times (E11).
Futurum Group CEO and analyst Daniel Newman said, per Liberty Times: "I think AMD is performing very well and could capture a 20% to 25% market share, which would translate into hundreds of billions of dollars in revenue" (E15).
| Metric | NVIDIA | AMD |
|---|---|---|
| Current data center GPU share (Futurum) | >95% | ~4.5% |
| Projected AMD share ceiling (Futurum/Newman) | — | 20%–25% |
| Top-10 AI firms running Instinct GPU workloads | — | 8 of 10 |
Baptista Research analyst Ishan Majumdar said the expanded AMD-Microsoft partnership marks an important commercial endorsement, showing AMD's ability to deliver integrated AI infrastructure rather than competing with NVIDIA on a single chip alone. However, he noted the partnership mainly reflects strategic validation and guidance on future revenue streams rather than an immediate, quantifiable earnings contribution, according to TechNews (E6).
Morningstar analyst Brian Colello said he expects AMD to disclose more partnership details at its Advancing AI event on July 23, and that AMD could even announce Anthropic joining its customer list, per TechNews (E7).
The two outlets' figures line up consistently: both TechNews and Liberty Times, citing Futurum research, put NVIDIA's current data center GPU share above 95% against AMD's roughly 4.5% — a gap AMD itself has not disputed. Yet AMD's own disclosure that 8 of the top 10 AI companies already run workloads on Instinct GPUs sits alongside Futurum's projection that Helios could lift AMD's share to 20%-25%, suggesting existing customer breadth is the basis for that growth estimate rather than a new customer base. At the same time, Baptista Research's Majumdar frames the Microsoft deal as "strategic validation" rather than a near-term earnings driver — a tempering note against Newman's "hundreds of billions of dollars" framing of the same 20%-25% share target. Whether that gap between validation and monetization narrows may become clearer at AMD's Advancing AI event on July 23, where Colello expects additional partner disclosures.
AMD said Helios is expected to begin shipping later this year (TechNews, E1), with Liberty Times specifying the second half of this year (E8).
Meta, OpenAI, Oracle, and Tata Consultancy Services (TCS) have all committed to deploying Helios, according to TechNews (E5).
AMD has not disclosed pricing, but Futurum Group estimated Helios at $5 million–$5.5 million versus $3.5 million–$4 million for NVIDIA's Vera Rubin, per Liberty Times (E13).
According to the Liberty Times and UDN, the White House pushed Apple, NVIDIA and SpaceX toward Intel after Washington took an roughly $8.9 billion, near-10% equity stake in the chipmaker. Deals followed, but Intel's foundry unit lost $10.4 billion over four quarters, and Apple still relies on TSMC for most custom chips.
According to United Daily News (UDN), Nikkei Asia reported that TSMC (台積電) has discussed raising 2027 wafer prices by a base 5%–10%, plus a 10%–15% surcharge on rush high-performance-computing (HPC) orders. Central News Agency (CNA) reports talks ran from June to July, effective early next year. TSMC told Nikkei it does not comment on pricing and calls its approach 'strategic, not opportunistic.'
According to a UDN Money report, TSMC (台積電) holds its earnings call on July 16, 2026, with analysts expecting about $56 billion in 2026 capital expenditure and over $150 billion across the next three years. Liberty Times Net reports TSMC guided Q2 revenue to $39.5 billion, up roughly 10% quarter over quarter.