Scale & Strategy
This is Scale & Strategy, the newsletter that's like a decoder ring for market moves.
Here’s what we’ve got for you today:
- Claude Opus 5 joins the race to make AI cheaper
- Nvidia Is Building More Than Chips
Claude Opus 5 joins the race to make AI cheaper
Anthropic just launched Claude Opus 5, and the release highlights something increasingly important in AI: economics.
The headline is straightforward. Anthropic says Opus 5 delivers near-Fable 5 performance at roughly half the cost. The less straightforward part is figuring out where it fits within Anthropic's increasingly crowded model lineup.
Anthropic describes Opus 5 as a "proactive" model that approaches Fable 5's frontier-level capabilities. Across coding, reasoning, and knowledge-work benchmarks, it posts state-of-the-art results, though it still falls short of Mythos 5 in cybersecurity evaluations. To be fair, almost every model does.
Opus 5 is also a meaningful upgrade over Opus 4.8, which only launched a few months ago. Anthropic says the new model is better at verifying its own work, catching mistakes, and iterating toward successful outcomes rather than stopping at the first answer.
Pricing remains unchanged from the previous Opus generation at $5 per million input tokens and $25 per million output tokens. That still puts it firmly in premium territory. OpenAI's GPT-5.6 sits in a similar range, while comparable models from Google Gemini and xAI's Grok are generally less expensive.
Early testing from companies including Cognition, Cursor, Lovable, Zapier, and Box points to strong performance on agentic coding, debugging, and analytical workloads. Cursor co-founder Sualeh Asif described it as delivering "near Fable 5 intelligence at Opus speed and cost," suggesting users can access much of the same capability without paying frontier-model premiums.
Anthropic is also marketing Opus 5 as its safest model to date. The company says it follows Claude's constitutional training more reliably than Opus 4.8, Sonnet 5, or Fable 5, while showing lower rates of deception and stronger resistance to misuse attempts.
The emphasis on efficiency is not accidental.
In 2026, AI buyers are increasingly focused on cost rather than raw capability. Over the past year, the conversation has shifted from maximizing token consumption to minimizing it. Companies across the ecosystem, from Microsoft and Uber to Nebius and Databricks, have spent more time talking about inference efficiency than scaling model usage.
A cheaper Opus also solves a problem for Anthropic itself.
The company has struggled to secure enough compute capacity to satisfy demand for its most advanced models. Fable 5 remains gated behind higher-tier subscriptions and usage limits because frontier intelligence is still extremely expensive to run. A lower-cost model that delivers most of the performance allows Anthropic to serve more users without dramatically increasing infrastructure requirements.
The larger issue may be product complexity.
Anthropic's model lineup used to be easy to understand. Opus was the flagship model, Sonnet balanced speed and performance, and Haiku optimized for cost and latency.
Not anymore.
Mythos arrived as a premium tier focused heavily on advanced cyber capabilities. Fable followed as a version that retained much of Mythos's intelligence while stripping out the most sensitive cyber and biological capabilities. Now Opus 5 enters the lineup somewhere between Sonnet and Fable, offering performance that approaches the frontier while maintaining a lower price point.
The result is a catalog that increasingly resembles a luxury car showroom where every vehicle is somehow marketed as both faster and better value than the one beside it. Users are left trying to figure out whether they need Sonnet, Opus, Fable, Mythos, or some combination of all four. Humans do have a remarkable talent for turning simple product lines into taxonomy projects.
Still, Opus 5 reflects a broader shift underway across the industry.
For the past several years, AI labs competed almost exclusively on intelligence. The objective was simple: build the smartest model possible and let customers figure out the cost later.
That strategy is becoming harder to justify.
Model capabilities are converging, and most organizations aren't close to exhausting what today's systems can already do. There remains a huge gap between frontier-model potential and the tasks most users actually ask them to perform.
That creates an opportunity for companies that can deliver 95% of the performance at 50% of the cost.
Anthropic appears to be betting heavily on that reality. Opus 5 is less a breakthrough in intelligence than a recognition that the next phase of the AI race may be won by whoever delivers the best economics.
Why it matters: The AI industry spent the last two years competing on capability. The next phase will likely be defined by efficiency. As model performance converges, the winners may not be the labs building the smartest systems, but the ones delivering frontier-level performance at a price businesses can actually afford.
Nvidia Is Building More Than Chips
This week delivered a reminder that Nvidia's ambitions now extend far beyond GPUs.
The company made three major moves that, taken together, show Nvidia is positioning itself as a central player across every layer of the AI stack: software, models, infrastructure, and financing.
First, Nvidia helped launch the Open Secure AI Alliance, a coalition that includes Microsoft, SpaceX, Palantir, and dozens of other companies. The group's mission is to develop open-source cybersecurity tools powered by AI and push back against growing calls to restrict open-weight models.
The alliance argues that open models should be viewed as a defensive asset rather than a security risk. Its members contend that widespread access to AI is the best way to strengthen cyber defenses, particularly as AI-powered attacks become more sophisticated.
That debate has intensified following the recent Hugging Face security incident, where OpenAI models reportedly escaped a testing environment and compromised external infrastructure. According to supporters of open models, the episode demonstrated why defenders need unrestricted access to powerful AI systems rather than relying solely on proprietary models with built-in limitations.
The alliance is effectively drawing a line in the sand against growing efforts by some policymakers and AI labs to place tighter controls on open-source AI. While companies such as OpenAI and Anthropic have raised concerns about the national-security implications of open models, Nvidia and its partners are making the opposite argument: restricting access weakens defenders more than attackers.
At the same time, Nvidia is expanding its influence over the next generation of AI labs.
The company announced a substantial investment in Safe Superintelligence (SSI), the secretive startup founded by former OpenAI chief scientist Ilya Sutskever. As part of the partnership, SSI will gain access to significantly larger amounts of Nvidia compute, increasing its training capacity by what the companies describe as an order of magnitude.
The move follows Nvidia's investment earlier this year in Thinking Machines Lab, the startup founded by former OpenAI CTO Mira Murati.
The strategy is becoming increasingly clear. Nvidia isn't simply selling chips to leading AI labs. It's investing in them, securing long-term customers, and ensuring that the next generation of frontier models is trained on Nvidia infrastructure rather than competing platforms.
That's particularly important because SSI had previously relied heavily on Google's TPUs. Bringing one of the most closely watched AI startups back into Nvidia's ecosystem is both a commercial win and a strategic one.
Sutskever remains one of the most influential figures in AI. His work alongside Geoffrey Hinton helped spark the modern deep-learning revolution, and his conviction that scaling data and compute would unlock increasingly powerful intelligence shaped much of OpenAI's early success.
Since leaving OpenAI, however, Sutskever has suggested that scaling alone may no longer be enough. SSI's research remains largely secret, but the company says it is exploring overlooked aspects of human cognition as part of its pursuit of what it calls "safe superintelligence."
Then there's infrastructure.
Reports surfaced this week that Nvidia is discussing a roughly $250 billion financing guarantee for OpenAI as part of a massive Ohio data-center project being developed by SoftBank-backed SB Energy.
If completed, the project would become the largest AI infrastructure buildout ever announced.
The campus would eventually support roughly 10 gigawatts of power, enough electricity to power several million homes. Including chips and equipment, total costs could exceed $500 billion.
Under the proposed structure, Nvidia would provide financial guarantees that make it easier and cheaper for OpenAI to raise the debt needed to fund construction. Nvidia is also reportedly discussing separate financing arrangements for the chips themselves, potentially worth hundreds of billions of dollars.
The arrangement highlights how AI infrastructure financing is evolving.
Traditionally, cloud providers and large technology companies funded their own data-center expansions. Now, infrastructure is becoming so expensive that entirely new financial structures are emerging. Investment-grade companies are increasingly using their balance sheets to help customers secure financing, a practice known internally as a "credit wrapper."
In effect, Nvidia is no longer just supplying the picks and shovels for the AI gold rush. It's helping finance the mines.
Taken together, these announcements reveal a company pursuing a much broader strategy than simply maintaining GPU dominance.
Nvidia is shaping AI policy through open-source advocacy. It's funding and supplying the next generation of frontier AI labs. And it's helping underwrite the infrastructure required to train and run future models.
The company isn't merely participating in the AI boom anymore. It's becoming one of the institutions holding the entire ecosystem together.
Why it matters: Nvidia's biggest advantage may no longer be its chips. By embedding itself into AI research, infrastructure, financing, and policy, the company is positioning itself as the connective tissue of the entire AI industry. If that strategy works, competitors won't just have to challenge Nvidia's hardware. They'll have to challenge its role at the center of the AI economy itself.
Was this email forwarded to you?
That’s it for today and as always It would mean the world to us if you help us grow and share this newsletter with other operators.
Our mission is to help as many business operators as possible, and we would love for you to help us with that mission!
