AI is getting worse as Google and Anthropic nerf AI models and limit usage

AI chatbot providers including Anthropic, Google, and OpenAI are facing growing user backlash as they raise prices, impose compute-based usage limits, and in some cases downgrade model capabilities. Anthropic reportedly more than doubled estimated per-developer daily costs, while Google is capping Gemini usage and charging credits for tools like Flow and Antigravity. The article argues the era of heavy AI subsidies may be ending, raising concerns about adoption, customer retention, and long-term business viability.

Analysis

This is less about consumer annoyance and more about the economics of inference snapping back to reality. The first-order effect is margin protection, but the second-order effect is demand destruction at the edges: casual users, small teams, and experimental workflows are the most price-sensitive, so they are the first to churn or revert to older tools. That creates a bifurcation where a few high-usage enterprise customers remain sticky while broad-based engagement weakens, which is bad for future model-improvement flywheels and product moat narratives.

The bigger strategic risk is that compute rationing slows adoption exactly when incumbents need scale to justify capex. If customers start treating AI like a metered utility instead of an always-on assistant, usage becomes episodic and harder to embed in daily workflows, which reduces switching costs and weakens pricing power over time. In that regime, the real beneficiaries are not the frontier labs but cheaper distribution layers, workflow software, and in-house/private-model deployments by large enterprises that can amortize usage over existing infrastructure.

For public comps, the bearish read is strongest on names where sentiment already prices in near-infinite TAM expansion and flawless monetization. Google and Microsoft face the risk that higher monetization friction slows consumer-to-enterprise conversion, while any evidence of model degradation can cap willingness to pay for premium tiers. Uber/Lyft are indirect reads-through: if AI usage economics disappoint, the market may get more skeptical of subsidy-then-monopolize strategies across tech, which compresses multiple expansion in adjacent platform stories.

Contrarian angle: this may actually be a disciplined normalization, not a demand collapse. If pricing is reintroduced gradually and enterprise bundles absorb usage, revenue per active user can rise even as headline complaints worsen, and the best-positioned incumbents can widen the gap versus smaller labs that cannot fund inference at scale. The key variable over the next 1-3 quarters is not user sentiment but churn/retention in premium cohorts; if that holds, the market may be overestimating the damage.

Key Decisions for Investors

• Short-term: buy 1-3 month downside protection on GOOGL and MSFT via put spreads into any post-announcement strength; thesis is that monetization friction and user backlash cap multiple expansion before fundamentals re-rate.
• Pair trade: long enterprise software with in-house AI adoption tailwinds vs. short AI-native consumer app exposure; the best relative winners should be workflow vendors that can absorb AI costs into existing seats rather than meter usage directly.
• Relative value: short LYFT vs. long UBER only if AI-related multiple compression broadens across platform/subsidy narratives; otherwise avoid making this a standalone thesis because fundamentals here are only second-order linked.
• For MSFT specifically, look for a 6-12 month call-spread hedge rather than outright shorting: if enterprise copilots remain sticky, the downside is limited, but a failure-to-scale premium AI usage can compress the premium embedded in the stock.
• If GOOGL sells off sharply on usage-limit headlines, consider scaling into a tactical long only after confirmation that ad/search engagement is unaffected; the risk/reward improves if the market overprices consumer AI disappointment relative to core cash flow durability.