Anthropic Goes on the Offensive: Opus 5 Retakes the Crown, AMD Ties Up a Mega-Deal
Lead: Anthropic launched its newest flagship model, Claude Opus 5, reclaiming the top spot on industry benchmarks — and in the same stretch, struck a computing partnership with AMD worth up to $5 billion and up to 2 gigawatts of capacity, reshaping the competitive landscape.
Claude Opus 5 reclaims the benchmark lead
On July 24, Anthropic released its new flagship model, Claude Opus 5, simultaneously across all platforms. According to reports, the model outperformed rival models that had previously held the top benchmark spots, swinging the generative-AI performance race back in Anthropic's favor. The industry has seen the lead change hands every few months in an unusually fast development race, and Opus 5 is the latest move in that contest.
A major AMD partnership, up to 2 gigawatts
Around the same time as the Opus 5 launch, Anthropic and semiconductor giant AMD announced a strategic partnership. AMD is reported to be making an equity investment in Anthropic worth up to $5 billion. At the same time, Anthropic plans to deploy AMD's newest AI accelerators — the Instinct MI450 series, packaged into data-center rack-scale systems called "Helios" — at a scale of up to 2 gigawatts to power its AI infrastructure. A gigawatt is an unfamiliar unit for most people, but it's roughly the output of a large power plant; committing that much capacity to a single AI company's computing needs is a striking figure. Reports indicate the first gigawatt of that capacity is expected to come online in the first half of 2027.
The deal has a second layer, too. AMD and Anthropic plan to work together on using Claude to accelerate workload optimization for AMD's GPUs and development of ROCm, AMD's software platform for GPU computing, and AMD itself is reportedly planning to roll Claude out broadly across its own internal design and engineering work. Some outlets describe it as a "circular" deal — one where money, hardware, and software expertise flow back and forth to reinforce both companies at once, rather than simple investment in one direction.
Why AMD?
NVIDIA GPUs have long been the default choice for AI computing, but Anthropic is reported to be deliberately diversifying its supply chain. According to reports, Anthropic is now working with AMD alongside NVIDIA, Amazon, and Google for chips and cloud infrastructure, and has secured more than 10 gigawatts of new computing capacity in total this year alone. Relying too heavily on a single supplier carries risk — from delivery delays to weaker pricing leverage — so spreading commitments across multiple partners helps secure the computing power AI development depends on. For AMD, which has been chasing NVIDIA in the AI chip market, landing a marquee deployment with a leading AI lab is seen as a significant win in its own right.
What it means for the competitive landscape
Anthropic and OpenAI sit at the front of the frontier-AI race, and the fight to lock down computing capacity has become just as intense as the fight over model performance. The AMD deal shows Anthropic isn't just competing on how smart its models are — it's also working to secure the computing foundation it will need for years to come. With a benchmark-topping model and a multibillion-dollar infrastructure bet landing in the same week, the AI race increasingly looks less like a pure contest of intelligence and more like a test of who can lock down — and sustain — the largest, most reliable supply of computing power.
Two announcements, one message
What makes the timing notable is that these were not two unrelated stories that happened to land in the same week — they reinforce each other. A new top-ranked model demonstrates that Anthropic's research team can still out-execute its rivals on raw capability. The AMD deal demonstrates that Anthropic has also lined up the physical infrastructure — chips, racks, and power — to keep training and serving models at that level for years, not just months. Taken together, reports frame the announcements as a signal to the rest of the industry, and to Anthropic's investors, that the company intends to compete on both fronts simultaneously rather than trading one for the other.
What comes next
The practical rollout will take time: the first gigawatt of AMD-powered capacity isn't expected online until the first half of 2027, meaning the computing benefits of the deal are still mostly ahead of Anthropic rather than already banked. In the meantime, the more immediate story is Opus 5 itself, and how long its benchmark lead holds before a competitor — most likely OpenAI or Google — answers with its own release. Given the pace the industry has kept up over the past year, few expect Opus 5 to hold the top spot for long, even as Anthropic's expanded computing base gives it a stronger platform for whatever comes after it.
Anthropic's New AI — and a Massive New Deal
Anthropic's new AI, Opus 5, just took back the top spot in performance tests. The company also announced a huge $5 billion computing deal with chipmaker AMD.
💡 The gist
- Anthropic released a new AI model, Claude Opus 5, and it topped the performance charts.
- Anthropic signed a huge deal with chipmaker AMD, worth up to $5 billion and up to 2 gigawatts of computing capacity.
- Building smart AI now depends just as much on having enough computing power as on the AI itself.
A new AI takes the top spot
On July 24, Anthropic released Claude Opus 5, its newest AI model, on every platform at once. On benchmark tests (think of these as standardized exams that measure how good an AI is at different tasks), it reportedly beat out the AI models that had previously ranked first. The AI industry sees the top spot change hands every few months in a fast-moving race between a handful of companies, and Opus 5 is Anthropic's latest move in that race.
A big deal with AMD
Around the same time, chipmaker AMD announced it would invest up to $5 billion in Anthropic. In return, Anthropic plans to build large-scale computing systems using AMD's newest AI chips (called the Instinct MI450, bundled into systems named "Helios") — up to "2 gigawatts" worth of them. A gigawatt is roughly the output of a large power plant, so this is a huge amount of computing power, dedicated specifically to running Anthropic's AI. The first portion of that capacity is expected to come online in early 2027. On top of the investment, the two companies also plan to work together on software: Anthropic will use its Claude models to help improve AMD's chip-related tools, and AMD plans to use Claude widely in its own internal engineering work.
Why work with more than one chip company?
Until recently, most AI companies relied mainly on chips from NVIDIA. Anthropic is now working with AMD as well as NVIDIA, Amazon, and Google, spreading its computing needs across multiple partners, and has reportedly locked in more than 10 gigawatts of new computing capacity in total this year. That way, if one supplier runs into shortages or raises prices, Anthropic isn't stuck depending on a single company. For AMD, meanwhile, landing a major AI lab as a customer is a big win in its own race to catch up with NVIDIA in AI chips.
Why it matters
The two announcements landing together sends a message: Anthropic wants to be seen as leading not just on how smart its models are, but on whether it has the computing muscle to keep building bigger ones for years to come. Increasingly, the AI race isn't just about building the smartest model today — it's also about who can lock down the most computing power to keep building tomorrow's.
The Smartest Robot Brain Yet — and a Giant Handshake Deal
Anthropic built a new super-smart computer brain called Opus 5. It is the best one yet. They also made a huge deal with a friend company called AMD.
What happened?
Anthropic built a new smart computer brain. It is called Claude Opus 5. People tested it against every other AI. It came out on top! Right now, it is the smartest one. 🏆 But being the smartest does not last long. It's like being the fastest kid at recess. Someone else is always training to beat you. That usually happens in just a few months.
What does that mean?
Smart AI brains need a lot of thinking power. The bigger the AI, the more chips it needs. Those chips hum away in giant buildings. So Anthropic made a deal with AMD. AMD is a company that builds chips. The deal has two parts. First, people say AMD is buying a piece of Anthropic. That could be worth up to $5 billion. It's like buying a small piece of a company. You do it because you believe in it. Second, Anthropic will fill its buildings with AMD's new chips. Running them all takes a huge amount of power. That's about as much as two big power plants make (2 gigawatts). And that $5 billion? It is a giant pile of money. If you saved $100 every single day, it would take you more than 130,000 years to save that much! Anthropic also stays friends with other chip companies. That way, it never runs out of thinking power.