Google is no longer running OpenAI and Anthropic to the same end. Competitors want AI that improves itself. Google wants an AI that understands the real world.
The split is easy to miss, because Google still ships models and still makes money. But his most recent release peaked at number 10 on the singles chart.
What Google Has and Didn’t Say About the AI Race
Google released Gemini 3.6 Flash on July 21. The drive was speed and cost, not raw power.
Follow us on X to hear the latest news as it happens
This version produces 17% fewer tokens than the previous version, Google’s blog he said. Symbols are small parts of the text that the AI writes. Fewer signs mean a cheaper solution.
Power is another matter. One published reading of the Artificial Analysis index ranked the model at 10. Every other major lab ranked above it.
Google never stops. It has already started its main training, for Gemini 4. A larger model, Gemini 3.5 Pro, is still being tested by partners.
Sundar Photosi has mentioned a different award. He has built way to personal assistants not a scoreboard success.
Advertisers are comfortable. Alphabet shares fell 6% in June after two majors researchers went to the competitors.
Inside DeepMind’s Bet on Global Species
A global example is AI that learns how physical objects behave. Gravity, motion, cause and effect. It predicts what will happen in a room, not the next word in a sentence.
DeepMind’s own website shows betting. It keeps the Genie 3 and Gemini Robotic under the global brand heading and has AI, which means software that controls the machine.
In May the lab expanded Project Genie to Street View. It also released SIMA 2, an assistant that learns to play inside 3D worlds.
OpenAI and Anthropic are aiming for something completely different. They want to repeat themselves, abbreviated as RSI.
In simple terms, this is intelligent AI to build the next, better AI. Then that one arrests another one later.
Author Alberto Romero argued on Tuesday that Google left the competition on purpose.
“Hassabis is betting on something else: global models. Models that can understand and model the real world, not just predict the next signal,” Alberto Romero. he wrote in the latest analysis.
Google didn’t say that. Demis Hassabis, who runs Google DeepMind, has never released RSI publicly.
Why a Rival Founder Says DeepMind Is an Outlier
A very sharp outside reading came from an enemy, months ago.
Jack Clark co-founded Anthropic. On May 4 he published story where AI is going.
He gave a 60% chance that AI could run its research by the end of 2028. He put 2027 at 30%.
Clark then asked which labs were chasing that goal. DeepMind, he wrote, “seems to be the most visible of the big three.” Circumspect means to be careful.
His evidence came from DeepMind itself. He also mentioned 2025 paper on AI security, written by co-founder Shane Legg.
Anthropic is very strong. It said in a renewing itself again and again Claude’s research has written more than 80% of the code that ships by May 2026.
Before February 2025, the share was close to zero.
The company has written AI is making better AI. In one run-through test, his brands won 52 times in April, against 2.9-year-old.
A skilled craftsman needs four to eight hours to make four times the profit on the same job.
Why Google Couldn’t Give Up on the AI Race Ever
Two points against the whole idea:
- Google leads the test that comes closest to measuring AI research capabilities.
MLE-Bench asks the model to generate automatic machine learning. The Gemini 3 sample scored 64.4% in February. That was a very good result at the time. Google ranked Flash 3.6 at 63.9% in July.
Retirement labs are usually not at the top of the board.
- Google is nowhere in the market.
Photsi told investors that the Gemini app has 950 million monthly users. Google jumped on it NVIDIA partnership for open AI this month. So is OpenAI and Anthropic.
Can Google Wait?
The case for patience is simple. Search ads pay for everything else, so DeepMind can take some time.
Lettering that indicates that the cushion is shrinking.
Revenue reached $119.8 billion in the June quarterup to 24%, according to results wrote July 22. The search alone brought $ 63.3 billion.
Then comes money. Alphabet poured $44.9 billion into data centers and equipment in three months. This is almost double the previous year.
The result was free cash flow of $5.86 billion. Free cash flow is what is left over after the building costs have been paid.
This figure was better than $10.1 billion in March. In December it was good for $24.6 billion.
Letters covered the gap by selling $49.6 billion of new shares in June. It borrowed another $20.3 billion.
Long-term debt doubled in six months, from $46.5 billion to $98.2 billion.
The line of accounts that shared AI research lost $5.79 billion, down from $3.37 billion. Patience now has a price.
What to Watch for in the Next 30 Days
- Whether the Gemini 3.5 Pro ships, that’s how it works
- Whether DeepMind shows global results linked to Gemini 4
- Whether or not Alphabet’s returns will be encouraging in September
- Whether Hassabis answers the self-promotion question directly
Gemini 4 is the real test. If global models work where coding stands, moving slowly seems wise rather than scary.
The next earnings report will show how long Letters will continue to pay for you to know.
A note Why Google May Quietly Leave the AI Race to OpenAI and Anthropic appeared for the first time BeInCrypto.
