The news
Microsoft (MSFT) and Meta Platforms (META) are steering employees away from Anthropic's Claude toward in-house tools, PYMNTS reported on October 5, 2026, citing The Information. In the same week, Barclays (BCS) announced a wider rollout of Claude across the bank. The contrast is a live lesson in AI vendor concentration: big buyers can shift their AI spending quickly, while others tie operations more deeply to one provider.
According to the report, Microsoft had been on track to spend $1 billion this year on internal use of Anthropic's technology and has since lowered that figure by one-third. A Microsoft spokesperson said the company has been steering staff to its own GitHub Copilot coding tool, though engineers can still choose other models, PYMNTS reported. Microsoft has reduced use of Claude Code, Claude models in Copilot and Claude Mythos, according to the report.
Meta reduced the number of employees using Claude Code from about 60,000 earlier in 2026 to 30,000 recently, PYMNTS reported, largely by pushing them to Muse Code, its Claude Code competitor, and MetaCode, an internal-only tool. Meta did not comment to The Information and did not immediately reply to PYMNTS' request for comment.
Barclays moved the other way. On October 1, 2026, Anthropic and Barclays said the bank plans to extend Claude Code, Anthropic's coding agent, to 50% of its developers by the end of 2026 and to most of its software engineers during 2027. The bank's Colleague Knowledge Assistant, built on Claude and live since 2025, has been adopted by more than 16,000 colleagues and handled more than 1 million searches, and Global Markets teams use Claude to process 120,000 emails a day, the announcement said.
Craig Bright, Barclays group co-chief operating officer, said in the announcement that software engineering and cybersecurity are being reshaped by increasingly capable AI systems. Paul Smith, Anthropic's chief commercial officer, highlighted the bank's plan to put Claude in the hands of most engineers by 2027.
The numbers
- Microsoft planned internal Anthropic spend (reported)
- $1 billion, since lowered by one-third
- Meta Claude Code users (reported)
- About 60,000 to 30,000
- Barclays Claude Code target, end-2026
- 50% of developers
- Barclays Knowledge Assistant users
- 16,000+
- Emails processed daily by Barclays Global Markets
- 120,000
Why CEOs should care
The Microsoft and Meta moves show the buyer's side of concentration risk working as intended: both had in-house alternatives ready, so they could cut usage when costs or strategy changed. Most companies do not. Before expanding any single model provider across coding, customer service and operations, CIOs should ask: if prices rise, service degrades or the vendor's policies change, how long would it take us to switch, and what breaks first?
For banks and other regulated firms, the Barclays plan illustrates how embedded one provider can become: an assistant used by more than 16,000 staff, email routing for a trading business, and coding tools for most engineers by 2027. CROs and boards should treat this like any critical outsourcing arrangement, with documented exit plans, tested fallbacks to a second model, and contract terms on notice periods, data portability and service levels.
CFOs should watch usage, not just licenses. PYMNTS also cited a Wall Street Journal survey finding that only 11% of about 400 businesses accurately forecast AI costs. Practical steps: build applications against an abstraction layer so models can be swapped, keep prompts and evaluation tests portable, and benchmark a second provider on the most critical workloads at least once a year.
The bigger picture
Anthropic is a leading supplier to U.S. businesses: PYMNTS cited the Ramp AI Index showing 43.5% of U.S. businesses paid for Anthropic subscriptions or tokens as of July. As the largest tech companies build their own coding tools, Anthropic's growth may lean more on enterprises like Barclays, while those enterprises take on more dependence on a single vendor.
What’s next
Watch whether other large tech buyers follow Microsoft and Meta toward in-house tools, how Anthropic prices enterprise renewals, and whether financial regulators ask banks to show exit plans for their AI model providers.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
