Sundar Gives Gemini the Costco Treatment After Four Straight Red Months (Greg Abel Approves)

“I’m tired, boss…” -Google shares lately

After shooting to the moon with a 46% rip across March and April, the landlord of the internet spent the next four months having what social media influencers might call a decidedly not-so-hot girl summer.

Alphabet just finished its longest monthly skid since 2015 after losing AI talent, rearranging DeepMind’s furniture, and making Wall Street wonder how the company that helped invent modern AI wound up chasing the school bus. 

Turns out, Sundar still has a few tabs open.

Google kicked off September Sell-Tember (our month-long holiday here at Stocks.News) by launching Gemini 3.8 Flash, its third Flash model in six weeks, alongside a cybersecurity sibling that can allegedly hunt down software vulnerabilities faster than any IT human alive.

The stock has gained about 1% since the announcement after dropping more than 1% during Tuesday’s bloodbath… meaning investors aren’t exactly “feeling moved” just yet.

But after four consecutive losing months, the world’s biggest spy agency will take a green candle wherever it can find one.

And the AI launch wasn’t Google’s only win.

A federal judge blocked the DOJ’s attempt to pry AdX from Alphabet’s cold, data-harvesting hands, leaving Sundar in possession of the ad machine regulators had spent nearly four years trying to repossess. 

So, yeah…

New AI models. Berkshire’s blessing. No forced ad-tech breakup.

We’re so back-ish.

“But is Gemini 3.8 Flash actually good, or did Google just add another decimal point?” - you, probably, avoiding the Teams meeting from Stall No. 2

According to Google, Gemini 3.8 Flash is its best reasoning and coding model yet, with major improvements in software engineering, multi-step problem-solving, and agentic tasks.

Which is conveniently what Scam Altman said about GPT-5 over a year ago before the internet tried to return it.

Still, Gemini 3.8 Flash brought receipts. It scored 54.9% on HLE-Verified, a benchmark spanning science, math, humanities, and professional fields. Google also says it beat most larger frontier models on a long-horizon software-engineering test while costing considerably less to operate.

And price is where Sundar appears ready to start throwing elbows.

Gemini 3.8 Flash costs 75 cents per million input tokens and $3.75 per million output tokens… the same introductory price as Gemini 3.7 Flash, despite the newer model’s improved capabilities.

Google is also rolling out pay-as-you-go pricing for Gemini Enterprise, discounts of up to 20%, monthly limits on agent spending, and a zero-dollar base subscription.

Translation: Sundar's entire gameplan appears to be smacking a Kirkland label on artificial intelligence and backing the forklift directly over Satya Nadella’s margins. 

Thing is, the warehouse economics might actually work. 

Nearly three-quarters of Google Cloud customers already use the company’s AI products, and Cloud CEO Thomas Kurian says those customers are spending 50% more than they originally committed.

The less bullish interpretation?

D.A. Davidson analyst Gil Luria says Gemini 3.8 Flash keeps Google in the race but probably doesn’t change its distant-third-place position behind OpenAI and Anthropic in the enterprise market.

Thank you, Captain Obvious.

On the other hand, Google’s cybersecurity model might be a more interesting flex.

Gemini 3.8 Flash Cyber was specifically designed to discover and patch software vulnerabilities. Google claims it delivers performance comparable to much larger frontier models while operating faster and at a fraction of the cost.

During internal testing, Chrome’s security team reportedly found that the model produced correct vulnerability patches at 2.6 times the rate of larger competing commercial models.

Another Google security team said it uncovered a critical foundational vulnerability in less than two hours… a process that would ordinarily require months of research. So take that for whatever it’s worth.

Meanwhile, Warren Buffett’s not-so-famous replacement delivered an Alphabet endorsement carrying more weight than an entire cubicle farm of analysts. 

Abel said Berkshire views Google as a significant AI player, partly because Berkshire can see how AI is being used (and delivering benefits) inside its own portfolio companies.

Of course, Google’s AI comeback won’t be cheap.

Alphabet is pouring enormous sums into data centers, chips, models, and other infrastructure, so Gemini will eventually need to generate enough enterprise revenue to justify the investment.

And that’s the real takeaway for investors.

Google may still trail OpenAI and Anthropic in parts of the enterprise AI market.

But Alphabet’s advertising empire remains intact, Cloud customers are spending more, Gemini releases are arriving at machine-gun speed, and Berkshire just stamped the company’s AI strategy with the Omaha seal of approval.

After four months of getting kicked down Wall Street’s staircase, Google finally started September with the railing in one hand…

And three new AI models, a federal judge, and Greg Abel in the other.

At the time of publishing this article, Stocks.News holds positions in Google as mentioned in the article.