ZDNET’s key takeaways
- Programming is AI’s killer app.
- The highest enterprise AI, particularly for programming, is Anthropic.
- Open-source AI is lagging behind its proprietary rivals.
For those who have been to ask J. Random Person on the road what the most well-liked enterprise AI Massive Language Mannequin (LLM) is, I wager you they’d say OpenAI‘s ChatGPT. As of mid-2025, nonetheless, Anthropic is the leading enterprise LLM provider, with 32% of enterprise utilization, in accordance with Menlo Ventures, an early-stage enterprise capital agency.
Earlier than you get too excited, although, understand that Menlo Ventures is a significant Anthropic investor. The agency has backed the corporate by a number of vital funding rounds, together with main their Sequence D spherical and taking part of their $3.5 billion Sequence E, which valued Anthropic at $61.5 billion.
Additionally: What happened when Anthropic’s Claude AI ran a small shop for a month (spoiler: it got weird)
In different phrases, Menlo Ventures has billions of causes to reward Anthropic. That mentioned, others additionally view Anthropic as the highest enterprise AI firm. As AI Journal put it, “Anthropic has established itself as the premier enterprise AI company by its Claude household of LLMs, reaching outstanding 1,000% year-over-year progress to succeed in $3 billion in annual recurring income.” Even by hyper-aggressive AI requirements, that is actual progress.
Behind Anthropic, you will discover OpenAI, which now has 25%; Google with 20%; and Meta Llama with 9%. All the way in which within the again, with a mere 1% you will discover DeepSeek, adopted by the remainder of the pack. Menlo Ventures credit Anthropic’s fast ascent to the robust efficiency of its Claude Sonnet and Claude Opus fashions.
(Disclosure: Ziff Davis, ZDNET’s guardian firm, filed an April 2025 lawsuit in opposition to OpenAI, alleging it infringed Ziff Davis copyrights in coaching and working its AI techniques.)
These numbers mirror the proportion of manufacturing AI use, not spending. They have been derived from a survey of 150 technical decision-makers at enterprises and startups constructing AI functions in the summertime of 2025.
What’s behind Anthropic’s success?
Three various factors are driving Anthropic’s rise. The primary is what Menlo Ventures calls “AI’s first killer app”: Code era. Whereas AI-created code quality remains questionable, however, extra builders are utilizing AI programming instruments than ever, and Claude has turn out to be programmers’ best choice with 42% of the market share. That is double OpenAI’s 21% share.
There are concrete examples of Anthropic improvement applications gaining reputation. For example, in only one yr, Claude helped remodel GitHub Copilot right into a $1.9-billion ecosystem. Claude Sonnet 3.5‘s 2024 launch confirmed how LLM breakthroughs could make attainable fully new classes reminiscent of AI IDEs, Cursor and Windsurf; vibe app builders, Lovable, Bolt, and Replit; and enterprise coding brokers, Claude Code and All Palms.
Another excuse Anthropic is successful is its use of reinforcement learning with verifiable rewards (RLVR) to coach its LLMs. Behind that sophisticated title lies a easy idea: You present clear, binary suggestions (1 for proper, 0 for incorrect) on the mannequin’s output. This works properly for programming AI instruments, the place the code both works or would not.
Anthropic has additionally led the way in which to LLMs that take step-by-step approaches to fixing issues and use exterior instruments to drag in knowledge to ship higher solutions. Briefly, Anthropic has been a frontrunner in creating AI agents. Apart from serving to individuals and programmers, this method may also help LLMs iteratively enhance their responses and combine instruments like search, calculators, coding environments, and different assets through the Model Context Protocol (MCP). This new open-source protocol allows LLMs and AI brokers to seamlessly join with the huge, ever-changing panorama of real-world knowledge, instruments, and providers.
Additionally: 7 strategic insights business and IT leaders need for AI transformation in 2025
That is essential as a result of Menlo Ventures additionally discovered that it isn’t value that drives firms to alter LLMs, it is efficiency. “This creates an sudden market dynamic: At the same time as particular person fashions drop 10x in value, builders do not seize financial savings through the use of older fashions; they simply transfer en masse to the best-performing one.”
This dynamic could change as soon as LLMs begin to mature and fashions start to succeed in comparable efficiency ranges. For now, although, as LLMs enhance massively from one launch to a different, firms are prepared to pay for the most recent and quickest.
AI within the enterprise
The examine additionally discovered that firms are steadily shifting from constructing and coaching fashions to inference, that’s, with fashions truly working in manufacturing. Startups are main the way in which, with 74% of builders now stating that the majority of their workloads are in manufacturing. Massive enterprises aren’t far behind, with 49% reporting that the majority or practically all of their computer systems are in manufacturing. Briefly, enterprises are actually utilizing AI, not merely constructing AI.
Lastly, the researchers mentioned that open-source LLMs have declined to 13% of AI workloads at this time from 19% six months in the past. The market chief stays Llama, albeit that Llama isn’t really open source.
Additionally: How agentic AI is transforming the very foundations of business strategy
Nonetheless, extra open-source LLMs have been showing. These embrace new fashions from DeepSeek (V3, R1), Bytedance Seed (Doubao), Minimax (Textual content 1), Alibaba (Qwen 3), Moonshot AI (Kimi K2), and Z AI (GLM 4.5) within the final six months.
They’re simply not used a lot. That is as a result of, regardless of their benefits, “higher customization, potential price financial savings, and the power to deploy inside non-public cloud or on-premises environments,” their efficiency has continued to “path frontier, closed-source fashions.” Add in that lots of the best-performing open-source LLMs to this point are from Chinese language firms that Western companies are cautious of, and open-source LLMs seem like stalling out.
Additionally: Anthropic tripled its revenue in 5 months – and this is why
So what does all this result in?
“Predicting the way forward for AI generally is a idiot’s errand. The market modifications by the week, with thrilling new mannequin launches, developments in basis mannequin capabilities, and plunging prices,” Menlo Ventures mentioned. Nonetheless, “circumstances are ripe for a brand new era of tolerating AI companies to be constructed on prime of at this time’s foundational constructing blocks.”
The query stays, nonetheless, “What is going to these foundational constructing blocks be?” OpenAI? Google, Meta? Anthropic? Keep tuned. We’re not but near with the ability to say which AI fashions will finally find yourself on prime.
Get the morning’s prime tales in your inbox every day with our Tech Today newsletter.
