Let's have a serious chat about Google

To conclude: The delay of Gemini 3.5 Pro is a negative signal, but it is far from enough to change my assessment of Alphabet's fundamentals. What truly warrants caution is not whether a specific model generation is delayed by a few weeks, but whether its capabilities continue to lag and product adoption faces obstacles, ultimately impacting Search, Google Cloud, and the developer ecosystem.

In fact, if you have been closely following 3.5 Pro, you would have noticed rumors in the community about a delay until the end of the month; I saw this five days ago. There are also reports that the latest checkpoint performance declined instead of improving.

I am primarily interested in 3.5 Pro for practical usage needs, rather than stock price speculation. I originally intended to use Gemini to refactor the entire UI of a project; based on my own experience, its visual and frontend output better suits my preferences. GPT-5.6 Sol performs well indeed, and despite frequent quota resets recently, its speed is undeniably slow.

Why do I believe a single delay is insufficient to shake the fundamentals?

First, Alphabet's most critical commercial engine remains Google Services, centered around search advertising, along with the rapidly growing Google Cloud. In Q1 2026, Google Search & other generated $60.4 billion in revenue, up 19% year-over-year; Google Cloud revenue reached $20 billion, up 63% year-over-year. This business model cannot simply stall because one Pro model is released a few weeks late.

Secondly, Google's AI strategy is not merely about selling model subscriptions but involves full-stack synergy: the infrastructure layer consists of data centers and TPUs, the middle layer features Gemini, and the top layer includes products and traffic entry points like Search, YouTube, Android, and Workspace. Its profit model and constraints are entirely different from those of OpenAI or Anthropic.

Of course, this does not mean Gemini's external competitiveness is unimportant. Gemini must serve Google's internal products while also helping acquire Cloud customers through APIs, Vertex AI, and Gemini Enterprise. Google doesn't need to rank first in every single benchmark, but it must maintain a leading position in the top tier over the long term.

At Google's scale, what truly matters is not just the upper limit of model capabilities, but the optimal balance of quality, latency, cost, and availability. Search, Ads, and Workspace face massive daily call volumes; prioritizing the Flash route is not a concession due to lack of capability, but a very pragmatic choice. I believe Google's strategic focus is highly correct.

The absolute cost advantage of TPUs is difficult to calculate precisely. Their more certain value lies in enhancing computing autonomy, reducing reliance on a single GPU supplier, and granting Google stronger supply chain bargaining power (which helps explain why TPUs won't significantly impact NVIDIA's fundamentals) alongside hardware-software synergy.

Therefore, my judgment is: the 3.5 Pro delay is a minor negative item worth noting, but it falls far short of being a fundamental turning point. What should be observed next are Search usage and monetization, Cloud growth and margins, enterprise and developer adoption rates for Gemini, and whether the massive AI capital expenditures will eventually yield sufficient returns.

As of now, I still believe Google is doing the right things and is on the path to executing them correctly.

I have confidence in them.

$Alphabet(GOOGL.US) $NVIDIA(NVDA.US) $Apple(AAPL.US)