If April was about AWS and Anthropic betting on each other with infrastructure, May was about Anthropic executing on it. A new frontier model, a developer conference that buried the lead on compute, meaningful updates to Managed Agents, and a funding round that closed the month at a scale few saw coming. In this issue, Alec MacEachern, VP of AI at UTurn Data Solutions, walks through what happened and what it means for teams building on AWS and Claude today.

Claude Opus 4.8 Lands on Bedrock

Claude Opus 4.8 became available on Amazon Bedrock on May 28, and the improvements are material for teams running production agents.

The model scores 69.2% on SWE-bench Pro, up from 64.3% on Opus 4.7, and is roughly four times less likely to let flaws in code it has written pass without flagging them. That last characteristic matters more in production than the benchmark number does. Agents that surface uncertainty and catch their own errors require fewer review cycles and produce more reliable output over long-running tasks. The 1 million token context window carries over from 4.7, and the model is available immediately on Bedrock, Claude Platform on AWS, Google Cloud Vertex AI, and Microsoft Foundry. Pricing is unchanged from Opus 4.7 at $5 per million input tokens and $25 per million output tokens.

Anthropic also launched Dynamic Workflows in Claude Code as a research preview on the same day, which allows teams to structure longer-running agentic work with more explicit control over how tasks are sequenced and resumed. For teams already on the Opus rate card, there is no pricing barrier to evaluating 4.8. The model ID switch is the only change required to start testing it against your existing workloads.

Code with Claude: The Conference, the SpaceX Deal, and Rate Limits

Anthropic held its first developer conference, Code with Claude, in San Francisco in early May. The agenda covered agentic development patterns, production agent design, and sessions with engineering teams from Asana, Cursor, GitHub, Replit, Vercel, and Spotify. The content was substantive. But the announcement that mattered most did not come from the main stage agenda.

Anthropic announced a partnership with SpaceX to allocate the full capacity of the Colossus 1 supercluster to Claude. For context, Colossus 1 represents over 300 megawatts of compute. The deal was a direct response to a problem that had become visible over the prior several months: demand for Claude, particularly from Claude Code users and teams running agentic workloads, had outpaced available capacity.

The immediate operational impact followed quickly. Rate limits were doubled across Pro, Max, Team, and seat-based Enterprise plans. Peak-hour limit reductions on Pro and Max were removed. API rate limits for Opus models were raised by as much as 17x in some tiers. For teams that had been designing around capacity constraints or throttling their agent call patterns defensively, that headroom is genuinely useful.

The conference is continuing with dates in London later this month and Tokyo in June. If your engineering team is building on Claude, the session content is worth tracking.

Claude Managed Agents Gets Self-Hosted Sandboxes and Private MCP Server Support

Also announced at Code with Claude, Claude Managed Agents now supports two capabilities that expand what is possible in enterprise agentic deployments.

The first is self-hosted sandboxes. Tool execution can now run in an environment you configure, whether that is your own infrastructure or a managed provider like Cloudflare, Daytona, Modal, or Vercel. The agent loop that handles orchestration, context management, and error recovery stays on Anthropic's infrastructure. This separation matters because it gives organizations control over where tools execute and what they have access to, without giving up the reliability and scalability of Anthropic's managed orchestration layer.

The second is private MCP server support. Managed Agents can now connect to MCP servers running inside your own environment. Combined with self-hosted sandboxes, this means both the tool execution environment and the services an agent reaches can run within your existing infrastructure boundaries.

For enterprise teams, this closes a gap that had made fully managed agents difficult to adopt. Agents can now operate with meaningful isolation and access control without requiring a custom-built orchestration layer to enforce it. We are actively evaluating how this fits into client engagements where data locality and tool access control have been sticking points.

Closing the Month: $65B and a New Frontier Model

Anthropic closed May with two announcements on the same day.

The Series H funding round brought in $65 billion at a $965 billion post-money valuation. That number reflects the pace at which enterprise and developer demand has moved. Anthropic reported run-rate revenue surpassing $30 billion earlier in the year, up from approximately $9 billion at the end of 2025. The capital will support continued infrastructure buildout, model development, and the global expansion the company has been accelerating throughout the year.

On the same day, as previously mentioned, Anthropic introduced their latest model, Opus 4.8. This launch follows Opus 4.7 by roughly six weeks, the shortest gap between consecutive Opus releases to date. That pace, combined with the compute deals signed in April and May, suggests the constraint on Anthropic's release cycle is shifting from model development to infrastructure capacity, and that the infrastructure is now catching up.

Scaling Quickly!

May confirmed what April's infrastructure announcements pointed toward. Anthropic is moving quickly, the underlying compute is being secured at scale, and the developer tooling is maturing in ways that are directly relevant to enterprise agentic deployments. For teams building on AWS and Claude, the roadmap is increasingly well-resourced and the operational tooling, from AgentCore to Managed Agents to Bedrock model access, is converging into something that supports production workloads rather than just pilots.

Alec MacEachern is Vice President of AI at UTurn Data Solutions, an AWS Premier Tier Services Consulting Partner based in Chicago. Over the past decade, he has held roles at NVIDIA, AWS, and Microsoft, helping organizations design, build, and scale AI solutions across a wide range of industries and platforms. Today, Alec brings that cross-platform experience to helping enterprises navigate cloud migration, modernize data foundations, and adopt production-ready generative and agentic AI solutions.