The Growing Gap in the AI Boom

A TechCrunch article published this past weekend under the headline The haves and have nots of the AI gold rush captures something many in the industry have been quietly discussing for months: the mood around the current AI boom is far from euphoric, even within the tech industry itself. The promise of widespread benefits is colliding with a far more selective reality.

The piece doesn't argue that the technology doesn't work. It asks who's actually benefiting, and who's being left behind.

Who are the "haves"

On the winning side, the pattern is consistent: large corporations with their own infrastructure, preferential access to cutting-edge models, established AI engineering teams, and budgets to absorb experimentation costs. These players can invest in deep integrations—custom MCP servers, bespoke agents, automated pipelines—and capture real efficiencies.

Then there's another category: companies that built their business model directly on Anthropic's, OpenAI's, or other providers' APIs. Product startups that have spent 18-24 months refining their offering and now have traction. For them, the boom is real and measurable in revenue.

Who are the "have nots"

The problem lies with the majority in between. Mid-sized companies that ran internal pilots without moving to production. Teams that spent months testing tools without a use case justifying the investment. Workers in sectors expecting reduced workload instead absorbing the learning curve cost with little training support.

There's also a layer of small businesses and freelancers with technical access to the same tools as the big players—Claude Code, API models, public MCPs—but lacking the time, technical judgment, or organisational context to turn that access into real competitive advantage. Democratised access doesn't equal democratised outcomes.

Why it matters now

This fracture matters for reasons beyond the headlines.

First, it feeds institutional scepticism. When middle managers at a mid-sized firm have spent a year hearing that AI will transform everything while their teams still battle the same bottlenecks, the pendulum swing—from hype to rejection—is almost inevitable. That closes doors to adoptions that would actually help.

Second, it pressures providers. Anthropic and the broader ecosystem have a direct stake in benefits spreading more widely: a polarised market is more fragile than one with broad adoption. It's no coincidence that recent months have brought moves toward cheaper models (Haiku 4.5), more accessible tools (Claude Code with its CLI interface and plugin system), and documentation aimed at smaller teams.

Third, it has concrete labour consequences. The job displacement narrative coexists paradoxically with AI talent scarcity. Both are true simultaneously, but for different populations. Those with skills to work with these tools are in high demand; those without face a transition with no visible safety net.

What's being done—and what isn't

Some governments and industry bodies are funding retraining programmes. But the pace of technological change far outstrips institutional training cycles. A six-month course designed in 2025 can be obsolete halfway through if the stack shifts radically.

On the private side, the most honest response we've seen comes from some integrators and specialist consultancies: openly acknowledging that most pilots don't reach production, identifying why, and charging for diagnosis before implementation. It's less flashy than "digital transformation", but it produces more honest results.

---

Our reading is straightforward: the AI ecosystem has solid tools and real use cases, but the value distribution problem won't solve itself with better technology. That TechCrunch names this directly at peak sector investment is, at minimum, a sign the conversation is becoming more honest.