Charles Edwards: Google Underestimated in AI, Gemini Once Led by 6 Months
Fund manager Charles Edwards stated that Google is overlooked in the AI field, having had the world's most advanced large language model six months ago. Since then, no new flagship model has been released, but the wave of AI adoption is shifting towards deep integration with existing products.
Edwards pointed out that the AI integration value in Google's products like Gmail, Chrome, and Search has not been fully realized. GOOG, as a leading AI pioneer, is worth holding long-term; if open-source models ultimately prevail, Google's network effects can still give it an advantage in a world where "intelligent costs trend to zero."
Gemini 2.5 Pro ranked first on the LMArena human preference leaderboard in May 2026, costing only 1/12 of Claude Opus 4, reflecting Google's competitiveness in both "performance + cost" dimensions.
Source: Public Information
ABAB AI Insight
Charles Edwards' views reflect the consensus of "valuation differentiation" in the AI investment circle of 2026: the market is overly focused on "new model releases" while underestimating the value capture ability of "product integration," similar to the valuation mismatch between "app stores vs operating systems" in the early 2010s mobile internet.
A comparable case can be seen in Microsoft's path from 2023 to 2025: the market initially questioned the integration speed of Copilot, but by 2026, the AI subscription rate for Office 365 reached 40%, proving that the monetization efficiency of "existing users + AI features" far exceeds that of "pure AI startups." Google's Gmail/Chrome/Docs combination has similar potential.
In terms of capital strategy, Google has chosen a three-phase strategy of "model leadership → product integration → ecosystem lock-in": after the release of Gemini 2.5 Pro at the end of 2025, it will maintain technological leadership, focus on deep integration of Gmail/Chrome/Docs in 2026, and is expected to achieve a closed loop of "AI features → subscription conversion → data flywheel" by 2027, forming a triple moat of "model + product + data."
Essentially, this represents a "value migration" within technological substitution: shifting from a "model performance competition" to a "product penetration competition." The underlying mechanism is that when the performance gap between models narrows to a point that is "imperceptible to humans," the competition focus shifts from "who is stronger" to "who is more ubiquitous," with Google's "product matrix + user base" becoming a decisive advantage.
ABAB News · Cognitive Laws
- Not releasing a new model for 6 months is not falling behind, but building momentum.
- The endgame of AI is not models, but penetration rates.
- When intelligent costs trend to zero, network effects become the moat.