OpenAI, Anthropic SDK Downloads Slow: AI Adoption Matures

Package downloads for OpenAI and Anthropic have slipped. Not dramatically, but enough to stand out: after 18 months of relentless double-digit growth, both companies are seeing their core SDK download numbers flatten or tick down. It's the first time AI-Buzz's proprietary tracking data has shown a market-wide cooling across the 50-plus companies in its index, and it raises a straightforward question about what's actually happening inside developer teams right now.

OpenAI at 193M downloads, growth nearly gone

The numbers are specific. According to AI-Buzz tracking data, OpenAI's PyPI package logged 193.6 million downloads over the past 30 days, up just 2% month-over-month. Its npm package was flat at 55.0 million. That's a company still moving enormous volume, but the growth rate has essentially disappeared.

For context, the prior 18 months saw consistent double-digit monthly gains.

Anthropic's numbers tell a similar story. Its PyPI package reached 61.2 million downloads, up 3%, and its JavaScript SDK held flat at 27.4 million, per the latest AI-Buzz tracking data. The parallel deceleration across two companies with different product strategies, customer bases, and recent release cadences, including OpenAI's GPT-4o launch covered by TechCrunch, makes a company-specific explanation hard to sustain. Something is shifting at the market level.

Metric	OpenAI	Anthropic
PyPI Downloads (30d)	193,604,545 (+2% MoM)	61,236,983 (+3% MoM)
npm Downloads (30d)	54,972,120 (0% MoM)	27,373,927 (0% MoM)
HN Mentions (30d)	1,152	2,335
GitHub Stars (Tracked Repo)	30,260 (openai-python)	2,915 (anthropic-sdk-python)

New installs slow when teams stop prototyping

Download counts are a reasonable proxy for new project starts. When a developer spins up a fresh environment, installs a package, and starts experimenting, that registers. When the same developer maintains and iterates on an existing production application, it largely doesn't. The most plausible reading of this data is that fewer developers are starting new AI projects from scratch, not that fewer are using AI tools overall.

That interpretation fits what engineering teams have been describing anecdotally for months: the proof-of-concept phase is winding down, and the hard work of running AI in production has begun. Production workloads favor stability over novelty. Teams lock dependency versions, reduce package churn, and stop pulling in new SDKs to evaluate. The download velocity naturally drops even as actual usage, measured in API calls, inference costs, and active deployments, may be growing.

There's a secondary factor worth considering. The pool of developers likely to experiment with generative AI tools has probably grown close to its near-term ceiling. Early adopters moved fast. The next wave is more deliberate.

That demographic shift alone would compress month-over-month growth rates even without any change in the underlying technology or developer interest.

Anthropic's 2,335 Hacker News mentions vs. OpenAI's 1,152

Downloads tell one story. Developer conversation tells another. Over the past 30 days, Anthropic registered 2,335 mentions on Hacker News (AI-Buzz data), compared to 1,152 for OpenAI (AI-Buzz data). That gap is striking given OpenAI's dominant install base, reflected in its openai-python repository sitting at over 30,000 GitHub stars versus Anthropic's 2,915.

Hacker News discussion skews toward architecture decisions, API design, and capability comparisons, exactly the conversations happening when teams are choosing which platform to build on for the long term. The volume of community discussion around Anthropic, including its API, suggests the company is winning the consideration phase even where OpenAI still dominates raw adoption. Whether that translates to download share shifts in coming quarters is the more interesting question.

What this means for investors and engineering leads

For investors, download growth has been a convenient signal of ecosystem momentum. That signal is now noisier. As the market matures, sustained usage, retention, and revenue per developer become more meaningful than new install counts. The companies that built large install bases during the experimentation phase now need to demonstrate they can convert those users into durable, paying customers running production workloads.

Engineering leaders face a more immediate decision. With fewer compelling reasons to keep evaluating new tools every month, the cost of switching platforms rises. Teams that haven't yet committed to a primary AI stack are under pressure to do so, because building production infrastructure around a moving target gets expensive fast. The companies with the most stable APIs, the clearest deprecation policies, and the most reliable uptime records are likely to consolidate share from here, regardless of which model scores highest on the latest benchmark.

The more telling data point will come in the next two to three months. If download growth stays compressed while API usage and revenue metrics climb, that confirms the maturation thesis: developers aren't starting fewer AI projects, they're just finishing more of them. If both download and usage metrics soften together, the picture gets more complicated.