05-07-2026

Anthropic just rented Elon's supercomputer

PLUS: ChatGPT lands in your spreadsheets, Gemma 4 gets 3x faster, and Samsung hits $1T

Good morning, AI enthusiasts. Anthropic just secured one of the most unexpected compute deals in recent memory — renting the full capacity of Elon Musk's Colossus 1 data center, a Memphis facility packed with over 220,000 NVIDIA GPUs.

The move immediately doubled Claude's usage limits for paid subscribers and introduced a new AI memory feature called 'dreaming' — but the bigger question may be what it signals for xAI if renting out infrastructure becomes its primary business model.

In today's AI recap:

  • Anthropic rents xAI's Colossus 1 supercomputer
  • Google's Gemma 4 open models get a 3x speed boost
  • ChatGPT lands inside Excel and Google Sheets for all paid users
  • Samsung hits $1 trillion on surging AI chip demand

Anthropic Just Rented Elon Musk's Supercomputer

From Larry Bruce:
"This deal is one of the more unexpected pairings in recent AI history — Anthropic quietly securing Elon Musk's own supercomputer cluster to power its next chapter. For developers and early adopters building on Claude, the immediate win is real: more compute means fewer limits and faster access when it counts."
— Larry Bruce, BDCbox

The Recap: Anthropic just took over the full compute capacity of xAI's Colossus 1 data center in Memphis — a facility housing over 220,000 NVIDIA GPUs — and CEO Dario Amodei unveiled the deal live at the Code with Claude developer conference, alongside a new AI memory feature called 'dreaming.'

Unpacked:

  • xAI had surplus capacity on Colossus 1 after moving its own model training to a newer facility called Colossus 2, which created the opening for Anthropic to rent the entire 300-megawatt data center.
  • Anthropic immediately passed the benefit to users — Claude Code usage limits for Pro and Max subscribers doubled overnight, and peak-hour throttling was removed entirely.
  • A new feature called 'dreaming' lets Claude Managed Agents automatically review past sessions on a schedule and store key patterns into memory, similar to how human sleep consolidates learning, making long-running AI workflows progressively smarter over time.

Bottom line: Anthropic now controls one of the largest GPU clusters in the world, giving it the raw compute to push Claude's capabilities forward at a faster pace. The bigger story may be xAI — if renting infrastructure becomes its core business, the company's identity as a frontier AI lab gets a lot more complicated.

Google's Open AI Models Just Got 3x Faster — With Zero Tradeoffs

From Larry Bruce:
"Speed gains on local AI models usually come with a catch — but Google's latest update to Gemma 4 breaks that pattern entirely. Developers and privacy-conscious builders running AI on their own hardware will want to pay close attention to this one." — Larry Bruce, BDCbox

The Recap: Google released Multi-Token Prediction (MTP) drafters for its Gemma 4 open-source AI models, delivering up to 3x faster responses on consumer devices — with no drop in output quality.

Unpacked:

  • The technique works by running a small helper model that predicts several words ahead during idle processing time, while the full Gemma model quickly verifies those predictions — generating text faster without changing how the model reasons.
  • On devices like Pixel phones and Apple M4 chips, users see 2.5x to 3.1x speed improvements depending on which model size they run.
  • The drafters are available now through popular frameworks including Ollama, MLX, VLLM, and SGLang under the Apache 2.0 license, so developers can use and build on them freely.

Bottom line: Faster local AI means you can run powerful models entirely on your own device — no cloud connection, no data leaving your hands. This kind of open, on-device performance pushes serious AI capabilities further into the hands of everyday developers and builders.

ChatGPT Is Now Living Inside Your Spreadsheets — For Everyone

From Larry Bruce:
"Spreadsheets have long been a barrier for professionals who lack advanced data skills, and ChatGPT's global rollout into Excel and Google Sheets changes that dynamic in a meaningful way. This is a practical productivity shift that early adopters and business teams can act on right now."
— Larry Bruce, BDCbox

The Recap: OpenAI has officially taken its ChatGPT integration with Microsoft Excel and Google Sheets out of beta, making it available globally to all paid subscribers. Users can now build, clean, and analyze spreadsheets using plain language — no formulas or advanced skills required.

Unpacked:

  • The integration is now live across every paid ChatGPT tier — Pro, Plus, Business, Enterprise, Edu, and K-12 — meaning most paying users can start using it today without waiting for any additional rollout.
  • Users can describe what they want in plain language and ChatGPT will build entire spreadsheet workbooks, clean messy data, explain confusing formulas, and generate dashboards or financial models like budget trackers and KPI reports.
  • Unlike ChatGPT's broad rollout, Microsoft's own Copilot integration for Excel remains enterprise-only, meaning small teams and individual users currently have no equivalent option on the Microsoft side.

Bottom line: This update lowers the bar for anyone who has avoided spreadsheets because the learning curve felt too steep. The ability to build a financial model or a launch forecast from a single sentence puts a genuinely useful tool in the hands of a much wider audience.

AI Chip Demand Just Made Samsung Worth $1 Trillion

From Larry Bruce:

"Samsung's trillion-dollar milestone is more than a market story — it's a direct reflection of how deeply AI infrastructure spending has reshaped the global economy. For developers and entrepreneurs building on top of AI, understanding what powers these systems under the hood matters just as much as the models themselves."
— Larry Bruce, BDCbox

The Recap: Samsung crossed the $1 trillion market valuation mark after shares surged more than 10% in a single day, fueled by AI's soaring demand for specialized memory chips and a blockbuster earnings report showing profits 8x higher than the same period last year.

Unpacked:

  • High-bandwidth memory (HBM) chips are specialized hardware that AI systems use to rapidly move massive amounts of data during training and inference — without them, modern AI clusters simply cannot run at the speeds they require.
  • Samsung's quarterly profits came in 8x higher than the same period last year, a surge that pushed shares up more than 10% in a single day and placed Samsung among an exclusive group of trillion-dollar companies that includes only one other Asian firm: TSMC.
  • Reports that Apple is exploring chip manufacturing deals with Samsung at US facilities added a second major catalyst to the valuation surge, signaling that Samsung's revenue potential stretches well beyond AI chip demand alone.

Bottom line: The AI hardware race is minting trillion-dollar valuations not just at AI labs, but at the chip suppliers powering them. Samsung's milestone signals that controlling the memory chips AI depends on may be just as valuable as building the AI itself.

The Shortlist

Other Top Stories

SAP announced plans to acquire 18-month-old German AI startup Prior Labs for $1.16B, betting on its tabular foundation models to power enterprise data workflows — the company also blocked the unauthorized AI agent framework NemoClaw from accessing its systems.

Apple agreed to pay $250M to settle a class-action lawsuit alleging the company overpromised on Siri and Apple Intelligence features that were delayed or never delivered, marking one of the first major legal consequences for AI feature hype.

DeepSeek revealed it is raising its first-ever outside investment round, with sources saying the Chinese AI lab could hit a $45B valuation — up sharply from the $20B figure floated just months ago, with Tencent and Alibaba among reported interested parties.

Snap confirmed its $400M deal with Perplexity AI has quietly ended, saying the two companies parted ways amicably — leaving Snapchat without the AI search partnership it had been counting on to compete in the agentic assistant space.

Get AI, CDP (Customer Data Platform) & SPA (Sales Process Automation) Tips

Get updates delivered
directly to your inbox.
2026 BDCbox© - All rights reserved

15-Minutes to 15 additional Sales a Month—Guaranteed.

Our conversational AI doesn’t just manage leads—it creates sales from your entire customer base. Let us show you how to squeeze more sales from your existing customer with a 15 minute, demo .

From Data to Deals: The Power of DIAA

Learn how Beatrix.ai transforms dealership data into personalized conversations that grow sales and service revenue. Schedule a 15 minute demo now.

15-Minutes to 15 additional Sales a Month—Guaranteed.

Our conversational AI doesn’t just manage leads—it creates sales from your entire customer base. Let us show you how to squeeze more sales from your existing customer with a 15 minute, demo.
  • Cass
  • NCOA
  • EmailValidation
  • EmailAppend
  • Mobile Phone Indicator
  • Mobile Phone Validation
  • Mobile Phone Append
  • VIN still owned and maintained by the customer