Alphabet reported second-quarter revenue above market expectations, but the earnings beat did not lift its shares. GOOG fell 1.24% and dropped another 2.92% in after-hours trading, leaving investors split over what the market was actually reacting to.
ABMedia said many observers pointed to two familiar explanations: macro uncertainty linked to an escalation in the Iran conflict, and Alphabet’s expanding capital expenditure plans. Evergreen Capital, a technology investor and researcher, offered a different read in a post on X. In its view, the bigger problem is that Gemini has fallen visibly behind in the race among top large language models, and that loss of confidence is what is weighing on the stock’s valuation.
Evergreen Capital says the issue goes beyond capex
The firm argued that higher capital spending is a headwind, but not the main reason sentiment has weakened. Its central claim is that Gemini’s performance has slipped relative to the leading AI models, and that the market is responding to Alphabet’s position in the AI race more than to the quarter’s headline revenue number.
Under that framing, a strong earnings print is not enough on its own. Investors are looking at whether Google can still hold a leading role in a market where frontier model capability increasingly shapes expectations for Search, Cloud and the company’s broader multiple.
Why Gemini may be lagging: DeepMind and TPU allocation
Evergreen Capital said Gemini’s weaker competitive position does not necessarily reflect a shortfall in technical talent. The researcher instead pointed to internal resource allocation, saying DeepMind has long had to compete with other Google business units for TPU compute. That, he said, has slowed progress on frontier model development.
He described that setup as hard to justify because DeepMind and Gemini sit at the technical foundation of Google’s broader commercial platform, yet were not treated as the clear internal priority in compute allocation. In his assessment, that amounts to a strategic mistake.
Sundar Pichai has recently addressed the issue in public, saying, “On TPU allocation, our number one priority is to make sure we have the frontier compute needed for AGI development. That is foundational to everything we do.” Evergreen Capital said it takes that commitment literally: DeepMind researchers should not have to compete with other Google teams for compute if they are pushing frontier work.
Two scenarios point to a valuation gap of more than 100%
Evergreen Capital laid out two scenarios to show how Gemini’s standing could shape Alphabet’s valuation over time.
If Gemini returns to the top tier
In the more optimistic case, Gemini regains a place among the leading AI models. The researcher said that could strengthen user stickiness in Google Search and support deeper advertiser commitment. Double-digit annual growth, in that scenario, would continue to build rather than slow from current levels.
Google Cloud would also benefit, he argued. Strong frontier model development could feed into a broader and deeper cloud services ecosystem, helping both revenue and gross margin.
If Gemini keeps falling behind
In the weaker scenario, Google’s revenue and earnings power would face steeper downward revisions over the next three to five years. As competitive risk rises, the price-to-earnings multiple the market is willing to assign would also contract.
The researcher quantified part of that gap. A move in the P/E multiple from the high double digits to the upper end of the 20s would imply roughly 50% of stock price downside. If the divergence in annual EPS compound growth rates between the two scenarios — described as a mid-to-high single-digit percentage-point gap — is added into the compounding effect, the difference in five-year total shareholder return could easily exceed 100%.
Why the timing matters more now
Evergreen Capital said the urgency around Gemini is tied to the direction of AI development itself. The industry, in its view, is moving toward automated research and recursive self-improvement. Once a lab pulls ahead in that environment, the advantage can persist longer because the lead reinforces itself. For those already behind, catching up becomes harder.
That is why the question of whether Gemini can rejoin the top rank matters so much for Google’s future business model, according to the researcher. He also pointed to Anthropic and OpenAI, which are actively building their own data centers. In his view, it is possible they could eventually offer hosted services that compete directly with hyperscale cloud providers, adding pressure to Google Cloud’s moat.
Still holding GOOG, with one condition
Even with those concerns, Evergreen Capital said it remains constructive enough to keep holding GOOG. The researcher said DeepMind is still one of the few top labs in the world with both dense talent and substantial compute, and that the team led by Demis Hassabis has not lost its underlying strength.
He also noted that Sundar Pichai has previously corrected course successfully in Alphabet’s history. But that confidence depends on execution. The TPU-priority commitment, he said, has to show up in verifiable action rather than remain an earnings-call statement.
Whether Gemini can return to the front rank within a reasonable period may be one of the biggest variables for Google’s valuation over the next five years.