> ## Content Index
> Fetch the complete content index at: https://wire.fourthweb.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Forrester's CEO Says 90% of AI Investments Will Fail
- URL: https://wire.fourthweb.ai/forresters-ceo-says-90-of-ai-investments-will-fail/
- Published: 2026-08-31T17:35:34.000Z
- Updated: 2026-08-31T18:00:46.000Z
- Description: The research firm that's been calling tech cycles since 1983 just put a number on AI's reality gap. Forrester Research CEO George Colony assessed the current AI investment wave as Nvidia commits $3.5B to MediaTek and Anthropic preps what could be a record-breaking IPO
- Author: Travis Wright
- Tags: AI Agent Economy, AI Infrastructure, Compute Wars, OpenAI, Anthropic, Google AI, Nvidia, IPO Watch

**The research firm that's been calling tech cycles since 1983 just put a number on AI's reality gap.**

### The Summary

- [Forrester Research CEO George Colony assessed the current AI investment wave](https://www.bloomberg.com/news/videos/2026-08-31/forrester-ceo-sizes-up-the-ai-boom-video?ref=wire.fourthweb.ai) as [Nvidia](https://wire.fourthweb.ai/tag/nvidia/) commits $3.5B to MediaTek and [Anthropic](https://wire.fourthweb.ai/tag/anthropic/) preps what could be a record-breaking [IPO](https://wire.fourthweb.ai/tag/ipo-watch/)
- Colony's firm has 40+ years of pattern recognition in tech hype cycles, giving his read unusual weight
- The timing matters: we're at the inflection point where AI moves from infrastructure build-out to actual business model validation

### The Signal

George Colony has seen this movie before. The Forrester founder watched the dot-com peak, the mobile revolution, the cloud migration, and the blockchain winter. When he talks about [AI hype cycles](https://www.bloomberg.com/news/videos/2026-08-31/forrester-ceo-sizes-up-the-ai-boom-video?ref=wire.fourthweb.ai), he's comparing datasets that span four decades.

The backdrop for his assessment is telling. Nvidia just wrote a $3.5 billion check to MediaTek, the Taiwanese chip designer. That's not R&D money or a strategic partnership announcement. That's betting-the-company-on-edge-AI money. MediaTek makes the processors that go into billions of phones, IoT devices, and smart TVs. If Nvidia is paying that price, they see the next wave of AI compute moving out of [data centers](https://wire.fourthweb.ai/tag/ai-infrastructure/) and into everything else.

> "The infrastructure bet and the application bet are diverging fast."

Meanwhile, Anthropic is queuing up an IPO that sources say could value the company north of $40 billion. They've shipped Claude, a genuinely useful model that's competitive with GPT-4 and [Gemini](https://wire.fourthweb.ai/tag/google-ai/). But they're still burning cash to train new versions, and revenue per parameter lags behind [OpenAI](https://wire.fourthweb.ai/tag/openai/)'s. The IPO timing suggests either supreme confidence or a need to reload the war chest before the music stops.

Colony's view cuts through both the maximalist hype and the doomer skepticism. Forrester tracks enterprise tech spending patterns. They see which CIOs are approving AI budgets and which are pumping the brakes. The delta between infrastructure spending (data centers, GPUs, model training) and application revenue (actual businesses paying for AI features) is the gap Colony is measuring.

Key splits in the AI economy right now:

- Foundation model companies burning $100M+ per training run
- Enterprise software adding "AI features" that are mostly autocomplete
- Real automation plays (code generation, data analysis, customer service) starting to show ROI

The Nvidia-MediaTek deal points to a thesis: the next trillion dollars in AI value comes from inference at the edge, not training in the cloud. Smart glasses that actually work. Phones that run models locally. Cars that process sensor data without round-tripping to AWS. That's a different bet than "transformers get bigger forever."

### The Implication

Watch what the research analysts who survived multiple hype cycles are saying versus what the VCs who've only seen line-go-up are saying. Colony's take matters because Forrester clients are the enterprises writing the checks. If they're seeing budget fatigue or ROI skepticism, it shows up in the data before it shows up in stock prices.

The divergence between edge inference bets (Nvidia-MediaTek) and foundation model bets (Anthropic IPO) suggests the market is fracturing. Some money thinks scale is everything. Other money thinks distribution is everything. Both can't be maximum right. For anyone building in this space, the question isn't "will AI win" but "which AI thesis wins first."

### Sources

[Bloomberg Tech](https://www.bloomberg.com/news/videos/2026-08-31/forrester-ceo-sizes-up-the-ai-boom-video?ref=wire.fourthweb.ai)