This was one of the most consequential weeks in xAI’s short history: the company formally ceased to exist as a standalone entity, folding into SpaceX under the SpaceXAI banner โ and the SpaceX S-1 IPO filing that followed revealed Anthropic is paying $1.25 billion per month to lease the entire Colossus 1 data center through 2029. Alongside the corporate drama, xAI shipped an aggressive agentic product stack: Grok Build (a terminal-native coding agent), Grok Skills (persistent instruction sets), a second wave of Connectors, and native OAuth access for SuperGrok subscribers inside third-party agents. Here is the xAI Weekly for May 27, 2026.
xAI Absorbed into SpaceX โ The “SpaceXAI” Era Begins
On May 6, Elon Musk announced that xAI will no longer operate as an independent company. Grok and X now function as the SpaceXAI AI division inside SpaceX โ a consolidation that followed February’s all-stock acquisition valuing xAI at $250B and SpaceX at $1T for a combined $1.25T. A small restructuring affected the Grok team (approximately 10 employees), and xAI’s separate corporate presence is effectively gone.
For enterprise buyers and procurement teams, the practical impact is limited in the near term โ API endpoints, billing, and Grok subscriptions remain under the xAI brand. But strategic planning should account for the shift: xAI is now a division of a company preparing for an IPO, with the compute infrastructure repurposed as an active revenue line.
SpaceX S-1: Anthropic Paying $45B for Colossus 1
SpaceX filed its S-1 IPO prospectus with the SEC on May 20 โ the first detailed public look at the combined SpaceX/xAI/X financial picture. The most striking disclosure: Anthropic has committed to paying $1.25 billion per month through May 2029, approximately $45 billion total, for exclusive access to the Colossus 1 data center (~220,000 NVIDIA GPUs, 300 MW).
xAI has moved its own training workloads to the larger Colossus 2 facility (550,000+ GB200/GB300 GPUs, ~1 GW), freeing Colossus 1 for the Anthropic lease. Musk framed the arrangement pointedly: “We reserve the right to reclaim the compute if their AI engages in actions that harm humanity.”
The S-1 positions Colossus infrastructure as a deliberate revenue center โ unused compute is being monetized, not mothballed. Watch for similar compute-leasing disclosures as the IPO process advances.
(TechCrunch, Business Insider)
Grok Build: xAI’s Coding Agent CLI
Grok Build launched in early beta on May 14 for SuperGrok Heavy subscribers โ a terminal-native coding agent with a planning TUI, clean diffs, up to 8 parallel sub-agents, worktree support, headless mode (-p), and Agent Client Protocol support for building orchestration apps on top of it.
The underlying grok-build-0.1 API model published May 19: 256K context window, scored 70.8% on SWE-Bench Verified โ competitive with the top tier of coding agents.
A notable ecosystem detail: xAI officially states Grok Build reads Claude Code marketplaces, plugins, MCPs, hooks, and instruction files (CLAUDE.md, .claude/rules/) natively, with zero configuration. The community is already swapping skills between both platforms, lowering switching costs for teams already invested in the Anthropic toolchain.
For engineering leads evaluating coding agents, Grok Build is now a credible third option alongside Claude Code and GitHub Copilot Workspace โ see our enterprise coding agent harness guide for an evaluation framework. The SuperGrok Heavy gate limits trial access for now โ but the SWE-Bench score and sub-agent parallelism are worth watching.
Grok Skills: Persistent Instruction Sets
Grok Skills launched May 18 for Grok 4.3 โ persistent instruction sets that survive across every conversation. Five built-in skills ship out of the box: Word, PowerPoint, Excel, PDF, and a Skill Creator. Custom skills are built by conversation or file upload and saved to a personal library.
Skills accept .zip, .skill, and .md files. The cross-compatibility with Claude Code skills and CLAUDE.md files means organizations with existing instruction libraries don’t need to rebuild from scratch.
For enterprise teams that have invested in prompt governance and instruction standardization, Skills provide a durable home for those assets inside Grok โ rather than relying on per-session pasting.
Connectors Wave 2 and Third-Party OAuth
Wave 1 Connectors (May 6) gave Grok read/write access to SharePoint, Outlook, OneDrive, Google Workspace, Notion, GitHub, Linear, and a Bring Your Own MCP surface. Wave 2 (May 22) added Vercel, Canva, Gamma, and S&P Global. Connectors are live on Grok Web, iOS, and Android.
Separately, on May 19โ21 xAI announced that SuperGrok and X Premium subscribers can authenticate into OpenClaw, Hermes Agent, and OpenCode via OAuth โ no API key required. This unlocks Grok TTS, Grok Imagine image and video generation, and Grok Build inside each platform.
For organizations evaluating Grok as an enterprise productivity hub, the Connectors surface is now broader than most competing assistant platforms at comparable price points. The community has already catalogued 14 documented connectors in the awesome-grok-connectors GitHub repo โ the typical signal that developer adoption is accelerating.
API: Eight Models Retired, Grok 4.3 Is the Unified Flagship
On May 15, eight model slugs were retired from the xAI API: grok-4-1-fast-reasoning, grok-4-1-fast-non-reasoning, grok-4-fast-reasoning, grok-4-fast-non-reasoning, grok-4-0709, grok-code-fast-1, grok-3, and grok-imagine-image-pro.
All deprecated slugs now redirect automatically to grok-4.3 โ but billing switches to grok-4.3 pricing: $1.25/1M input tokens, $2.50/1M output tokens. Teams that haven’t explicitly migrated may see materially higher costs on their next invoice.
Grok 4.3 is now the sole general-purpose model in the xAI API: 1M token context window, four reasoning effort levels (none, low, medium, high), and leading benchmark scores on agentic tool calling and instruction-following. Third-party analysts currently place it above or alongside GPT-4o and Claude Sonnet in structured output and multi-step agent tasks.
Action required: Audit your API usage. Any integration still hitting a deprecated slug should be updated to an explicit grok-4.3 reference before the next billing cycle. Our xAI and Grok enterprise buyer’s guide covers vendor assessment and migration planning.
(docs.x.ai migration guide, docs.x.ai models)
Aurora Image Generation: Quality Mode API
Grok Imagine Quality Mode (grok-imagine-image-quality) launched May 6 as the new recommended image model for enterprise developers, replacing the deprecated grok-imagine-image-pro. It delivers higher photorealism, stronger multilingual text rendering (Japanese, Korean, Chinese), and tighter prompt adherence โ up to 2K resolution, up to 10 images per request, 3 reference images for compositing, priced at $0.05/output image.
Grok Imagine Agent Mode (an AI-powered canvas for interactive image and video creation) is now available in chat โ currently gated to SuperGrok subscribers, not X Premium.
Federal Market: FedRAMP High and GSA OneGov
xAI is pursuing FedRAMP High authorization, sponsored by USDA and assessed by Schellman โ the certification required for Grok deployment in the federal government’s most sensitive workloads. Multiple agencies (USDA, DOE national labs, HHS, Treasury, OPM) are already using Grok in varying capacities. The multi-year process covers 400+ security controls.
The GSA OneGov agreement already provides all U.S. federal agencies access to Grok 4 and Grok 4 Fast at $0.42 per organization for 18 months through March 2027, with a dedicated xAI engineering team for implementation.
For government-focused consultancies and integrators, xAI’s federal footprint is expanding faster than most expected. FedRAMP High, once achieved, would open the highest-sensitivity federal workloads โ currently the exclusive domain of a small number of certified platforms.
Anysphere (Cursor) Strategic Deal
In April, xAI struck a deal with Anysphere โ maker of Cursor IDE โ giving xAI the option to acquire the company for $60 billion later in 2026, or to pay $10 billion for joint collaborative work. This positions xAI and Grok Build as direct competitors to the Cursor IDE ecosystem while simultaneously creating a pathway for consolidation.
A completed acquisition would rank among the largest AI tooling transactions on record and would directly challenge GitHub Copilot and the broader VS Code extension ecosystem. No acquisition has been announced, but the option window appears to run through the rest of 2026.
What to Watch
- SpaceX IPO timeline: The S-1 is filed. Watch for pricing announcements and the formal listing schedule โ it will crystallize the combined SpaceX/xAI/X valuation and surface additional detail on the Colossus compute-leasing strategy.
- Anysphere/Cursor acquisition: The option to acquire Cursor for $60B expires at some point in 2026. Completion would reshape the IDE and coding-agent landscape overnight.
- Grok 4.3 billing impact: The May 15 model retirements affect any team using deprecated slugs. If you haven’t audited your API integrations, do it this week.
- Grok Connectors on iOS/Android: xAI has stated full Connectors support is coming to mobile apps soon โ meaningful for enterprise users managing workflows on the go.
- FedRAMP High progress: Watch for assessment milestones from Schellman and additional agency adoption announcements, particularly from DOE and HHS.
- Grok 5 / Colossus 2: Community signals point to a potential 6-trillion-parameter architecture targeting late 2026, trained on Colossus 2’s 550K+ GB-class GPUs. No official confirmation, but the infrastructure scale is consistent with those ambitions.
That is this week’s xAI Weekly. The SpaceXAI consolidation and Anthropic’s $45B compute commitment define the strategic moment โ xAI’s infrastructure is now a business in its own right, not just a cost center. Grok Build, Skills, and Wave 2 Connectors collectively represent the most aggressive agentic product push xAI has made to date. And the API model retirement sweep means some teams have an immediate billing audit on their to-do list.
If your organization is evaluating Grok Build for coding workflows, planning a federal AI deployment, or navigating the API migration, book a strategy session with Big Hat Group โ we specialize in helping enterprise teams navigate exactly these decisions.
Check back next week for the latest.