Tech Brew keeps business leaders up-to-date on the latest innovations, automation advances, policy shifts, and more, so they can make informed decisions about tech.

Tech Brew

Big Tech

Patrick Kulp

techbrew.com

When it comes to AI, Big Tech has to spend money to make money (eventually…fingers crossed).

Companies like Microsoft, Google, and Meta have gotten that first part down, pouring 

 into new data centers and other AI infrastructure. Microsoft and Google saw those bets start to pay off nicely in the most recent quarter, with major growth in their respective cloud businesses. Meta also claimed its strong ad sales were fueled by AI efficiencies.

Investors cheered strong results from Meta and Microsoft this week with more than $500 billion in stock market gains, according to 

 Microsoft’s Azure cloud business pulled in $75 billion in the past year, notching 34% YoY growth. Along with strong overall earnings and revenue beats, the results fueled the tech giant’s 

 as the second-ever $4 trillion company this week, behind Nvidia. This was the first time Microsoft broke out revenue for Azure specifically, beyond the “cloud services” bucket.

Microsoft has said it would slow down capital expenditures in the next fiscal year, which started this month, but it still plans to spend $30 billion this quarter. CFO Amy Hood said the company is trying to keep up with a backlog of existing demand.

Microsoft said its Copilot apps now have 100 million users every month.

 While Microsoft plans to slow spending, Google said last week that its AI infrastructure spending would be $85 billion for the year, $10 billion more than previously expected, because of strong cloud demand. Google also marked a banner quarter for its cloud arm with 32% YoY growth, as well as broader earnings and revenue beats.

Alphabet recently notched a win when OpenAI said it would add Google Cloud to its roster of suppliers for ChatGPT, as OpenAI continues to diversify beyond its partnership with Microsoft, 

 Then there’s Meta, where strong earnings and revenue beats ensured that Mark Zuckerberg was able to spend most of the call discussing his favorite topic: the company’s new quest for superintelligence. Meta has spent recent weeks setting its game board with 

 organized around former Scale AI CEO Alexandr Wang.

All those massive pay packages for AI researchers will make employee compensation Meta’s second biggest cost driver for next year, beyond AI infrastructure.

In a missive Zuckerberg shared ahead of earnings, the Meta CEO said “novel safety concerns” will mean the company has to be “careful about what we choose to open source.” Meta’s Llama model family has been among the leading options for 

He clarified in the call that while “I don’t think that our thinking has particularly changed” on open-source, some models might be too big or unsafe to open-source in the future. “We’re getting models that are so big that they’re just not practical for a lot of other people to use. So it’s, we would kind of wrestle with whether it’s productive or helpful to share that or if that’s really just primarily helping competitors or something like that,” Zuckerberg said.

Google, Meta, Microsoft on 3 pillars surrounded by binary code

The cloud business is booming for Microsoft and Google

During an earnings call, Mark Zuckerberg revealed more about Meta’s quest for superintelligence.

No, it’s not a TikTok aesthetic that involves dressing like a spy: AgentCore is a new toolkit from Amazon’s cloud arm designed to help companies build and deploy 

 in AI, companies everywhere are itching to 

 to full-fledged production can be daunting.

Amazon Bedrock AgentCore is the latest in a 

 of new products from tech giants aimed at easing that process. The platform has tools for managing memory, observability, helping agents access other digital services and the web at large, running generated code, and coordinating between agents.

Amazon announced AgentCore, along with a new agent marketplace and a 

, at its AWS Summit in Manhattan last week.

“There are a bunch of missing pieces that make it really difficult [to] move from [proofs of concept] that are sitting in laptops to agents in real-world production use cases,” Swami Sivasubramanian, VP of AWS Agentic AI, said on stage at the event.

The hype around agents seems as strong as ever at the moment, and enterprises have plenty of choices when it comes to these kinds of tools. 

, and other tech companies have their own agentic builders for customers looking to delegate to so-called “

Preethi CN, tech executive and advisor for agentic AI at AWS, told Tech Brew that one of the key things that sets AgentCore apart from this crowded field is flexibility. All of the component services are independent, CN said, and customers can “pick and choose” between them as needed. Companies are also not tied to Amazon’s own Nova foundation models; they can choose any of the catalog of models that Bedrock offers.

That kind of flexibility might be useful for a customer like Intuit, which has already built many of these agentic tools on its own through an internal tool called GenOS, according to Intuit SVP and chief data officer Ashok Srivastava. He said his team will explore these new AgentCore tools and decide if and where they might make sense for the new agents the financial software giant 

“When we built the agents that we launched a few weeks ago, a lot of these technologies weren’t available, and so we built our own,” Srivastava told Tech Brew. “As [cloud providers] build these capabilities out—if it makes sense—we can adopt them into our platform.”

Meanwhile, Amazon has also assembled an agent R&D lab in San Francisco built around its quasi-acquisition of agent startup Adept last year. The Amazon AGI SF Lab’s goal is to “develop new foundational capabilities for enabling useful AI agents that can take actions in the digital and physical worlds.”

Kelsey Szot, former co-founder and president of Adept, who’s now a member of the technical staff at the AGI SF Lab, said the lab is dedicated to research questions that lead to end results for AWS customers. The lab recently yielded a model called Nova Act, which powers agents to complete tasks within web browsers.

Szot said reliability has been the biggest obstacle for companies taking agents from experimentation to production—but that’s finally changing.

“Unfortunately, in the field, the last few years have seen a ton of demos and not a lot of real results of actual production deployments that are having business impact in the agent space,” Szot told Tech Brew. “It’s only now that we’re seeing the level of reliability that we need to actually make this possible.”

Amazon jumps onto the enterprise agent hype train

AWS’s new toolkit aims to help companies move past proof of concept.

In the high-stakes war for AI talent, you never know which way the wind will blow next.

 been in talks to acquire AI coding platform Windsurf when Google 

 to hire away the startup’s top talent and secure nonexclusive licensing rights to some of Windsurf’s assets. Then Cognition, the AI startup behind coding agent Devin, 

The episode is the latest in a string of reverse-acquihire-style deals that have become fashionable in the AI space. A Big Tech company will peel off top employees from a startup and license rights to some tech assets, while stopping short of an outright acquisition. Some examples include 

 These types of deals notably sidestep the traditional merger review process, which might have been especially attractive under the Biden administration’s 

 pointed out that these deals have another advantage that could explain why they’ve lived on into 

: expediency. In a breakneck AI development race where every week counts, absorbing startups this way allows for much quicker integration than a traditional acquisition.

“We note this deal mirrors the increasing[ly] popular partnership structure between large technology companies and some fast-rising GenAI startups that upgrades talent and provides access to key new GenAI products, without facing the scrutiny of equity ownership or acquisition,” Colin Sebastian, senior research analyst at financial firm Baird, wrote in a research note.

While Cognition ended up purchasing the remains of Windsurf, there were a couple of days when it seemed that remaining employees might be left in the shell of a company without the usual trappings of a startup exit.

The situation raises the question: If there’s a chance that a startup’s founders and other top employees might jump ship en masse, does that change the incentive math for tech pros considering working at one?

Rudy Torrijos, PitchBook’s director of industry research, told Tech Brew this particular deal was the result of a one-of-a-kind situation that’s not likely to become commonplace. However, it could lead venture capitalists and startup leadership to carefully consider internal structures and ensure that incentives are aligned for companies.

“This will likely prove to be a rare occurrence and was likely due to very unique circumstances,” Torrijos said in an email. “However, this event is likely to spur VCs and startup boards to review founder [and] CEO compensation or employment agreements and make sure leadership interests are aligned with investors and employees more explicitly and transparently.”

Putting structures in place that discourage founders from abandoning teams is especially important in times of major transformation in the tech industry, Torrijos said.

“In times of significant platform changes, like during the rise of the PC and the dot-com era, leadership matters most, suggesting that VCs and startup boards will likely add provisions, incentives, or disincentives for abrupt CEO departures,” Torrijos said.

 The episode also shows just how heated the battle over top AI talent has become among leading AI companies. At a time when Meta has been offering OpenAI researchers 

, big-name AI researchers are currently among the most in-demand resources in the AI race.

“Larger companies are desperate to hire anyone with raw talent or extraordinary experience,” Torrijos said. “Advanced technical talent has always been hard to find.”

Image of people shaking hands on a multicolored surface.

Windsurf drama shows how poaching is a new norm for AI deals

Google snagged execs from the AI coding platform before another startup bought the rest of Windsurf’s assets.

That’s what two California judges recently ruled about Meta and Anthropic’s use of certain copyrighted books in training their respective LLMs. The companies each faced a lawsuit from a different set of prominent authors.

These are the first major rulings to weigh in on 

 of whether it should be legal for developers to 

Tech companies argue that falls under the fair use doctrine, a legal exception that allows for nonpermissioned use of protected works in certain contexts, like news reporting, research, or parody. But copyright holders like authors, news publishers, and other media creators tend to argue that AI models harm the market for their work or aren’t sufficiently transformative in output, both of which would mean fair use isn’t applicable.

These arguments are playing out over a slew of 

 currently in the court system. Lawyers have told us that it will likely take a Supreme Court decision to settle this for good—and that could take years.

But for now, the Anthropic and Meta rulings offer a window into how judges are thinking about these questions. And so far, they’ve sided with AI companies, albeit with major caveats.

While a judge ruled earlier this year in favor of Thomson Reuters against legal AI startup Ross Intelligence in an AI copyright decision, the case notably did not involve 

 AI. And lawyers told us there were some other factors that made it less broadly applicable.

Authors Andrea Bartz, Kirk Wallace Johnson, and Charles Graeber sued Anthropic for the use of their books—both pirated and legally purchased—in the company’s “central library,” a hub of millions of works that Anthropic pulled from to train various LLMs.

Senior US District Judge William Alsup ruled that training Claude on these books was “exceedingly transformative,” thus qualifying as fair use.

Alsup also dismissed the idea of economic harm caused by Claude flooding the market. “[The] complaint is no different than it would be if they complained that training schoolchildren to write well would result in an explosion of competing works,” he wrote in the ruling.

Alsup did, however, grant that Anthropic’s downloading of millions of pirated books for its central library was “theft” and allowed for a trial on that matter specifically.

 Advocates for copyright protections might be heartened to read the first three pages of District Court Judge Vince Chhabria’s ruling on 

 between a group of high-profile authors and Meta.

In “most cases,” non-permissioned training on copyrighted works is likely illegal, he wrote. Copyright law is about preserving the incentive for human beings to create works of value, and a flood of AI-generated media will “dramatically undermine the incentive for human beings to create things the old-fashioned way,” Chhabria wrote.

Unfortunately for these specific authors, including Sarah Silverman and Junot Díaz, theirs are not “most cases.” The judge said the two main arguments their lawyers presented—that Meta’s Llama can reproduce snippets of their books, and that training without permission harms their ability to license their work to AI companies—are “clear losers.” A more persuasive argument would’ve centered on market dilution, but the plaintiffs “barely give this issue lip service.”

Chhabria sought to narrow the ruling to the details of this specific case and these specific authors.

“This ruling does not stand for the proposition that Meta’s use of copyrighted materials to train its language models is lawful,” the judge said. “It stands only for the proposition that these plaintiffs made the wrong arguments and failed to develop a record in support of the right one.”

Big Tech scores major wins in the battle over AI copyright

Meta and Anthropic each prevailed in “fair use” court cases against groups of book authors.

You may have noticed a new tab appearing alongside the familiar news, images, and shopping options on the Google homepage: AI Mode.

It’s the tech giant’s latest answer to 

 from AI apps like ChatGPT and Perplexity, and the next step after the AI Overviews it now inserts atop many results pages. AI Mode taps Google’s Gemini models to offer a more conversational search experience while drawing from content around the web.

Google started testing the feature in March, rolling it out to users nationwide in recent weeks.

 people access information on the web. With AI Mode tucked beside the familiar lists of links, it could accelerate these changes. We spoke with Soufi Esmaeilzadeh, director of product management for Google Search, about what this might mean for the future of search.

 Esmaeilzadeh said Google has been encouraged by how AI Overviews have led to more complex and nuanced queries. AI Mode is designed to build on that with an under-the-hood feature called “fan outs” that breaks questions into subtopics for the model to explore.

“It brings forward a comprehensive response that gives users an overview and then enables them to go deeper themselves when they want to go into various websites and dig in further,” Esmaeilzadeh said.

But will searchers actually bother to dig further into websites? That’s a key concern for companies that rely on 

 traffic from Google searches. The trade group News/Media Alliance, for instance, has 

Esmaeilzadeh said “bloggers, creators, and web content” are critical to the success of AI Mode. She claimed AI Overviews have yielded higher-quality clicks compared to traditional search results because users are more inclined to spend time on the site.

“The better we understand the intent, the better we can get people to the best content of the web that’s relevant for them. And then also, more importantly, as we open up more and more intents, more and more questions, we are growing the kinds of things people are asking for, and then we’re able to do more and more connections to the web given that,” Esmaeilzadeh said.

“That is something that is continually central and important to us. And I think it’s important to our users; they want to connect to the rest of the web.”

 Esmaeilzadeh demonstrated how a searcher might use the tool to conduct an in-depth survey of nearby restaurants. AI Mode might respond to more detailed requests and surface details like signature dishes. A searcher could then follow up with questions, asking which of the menu options are kid-friendly, for instance.

She said Google is better equipped to provide access to real-time information than newcomers like OpenAI because of the years it’s spent assembling databases on shopping, local businesses, and reviews.

Does AI Mode change how those businesses need to think about making that information available online? Because the underlying model, Gemini 2.5, is essentially performing Google searches, the concerns for businesses are the same, she said. But because it “fans out” with those subtopic-based searches, AI Mode should pull from a wider swath of content, Esmaeilzadeh said.

Google has done many experiments with generative AI search at this point, from the early days of Bard to AI Overviews. The goal of all of them, Esmaeilzadeh said, has been to expand access to information.

“Ultimately, we want people to be able to ask anything that’s on their mind, to make it easy for them to ask their questions, to make these follow-ups and continuation easy,” she said.

An image of Google's AI mode on a phone screen

Google search goes AI Mode

The tech giant is moving beyond AI overviews. What could it mean for the future of search?

It’s peak developer conference season, and AI announcements are flying in fast.

Google, Microsoft, Anthropic, and OpenAI have all elbowed each other to release news of their latest models, partnerships, and tools as the race to own generative AI remains as intense as ever.

 that can act beyond the realm of a chatbot—whether they’re helping consumers shop or book tickets, or workers code or make sales.

But companies diverged around how much emphasis they put on AI for use at home versus at the office. The Economist 

 some companies may be growing disillusioned with offerings in the latter category.

Here are some recent highlights from the AI world:

Anthropic, which has carved out a lane for its safety consciousness, released Claude Opus 4 and Claude Sonnet 4. But it was details in the release notes that grabbed the most attention: In safety testing, Opus 4 tried to blackmail people it believed were going to shut it down. In another test, Opus 4 tried to whistleblow on a hypothetical pharmaceutical company committing fraud 

 regulators and news outlets including investigative site ProPublica. The company is thus releasing the model with 

 to buy former Apple designer Jony Ive’s startup for $6.5 billion in a coupling that could jolt the race for AI-enabled hardware. There’s already been plenty of speculation about the assistant device OpenAI is purportedly building with Ive’s design prowess.

After spotlighting consumer products at its 

, Microsoft is once again locked in on all things enterprise and agents. The company let loose a barrage of 50 AI tool announcements meant to position Microsoft at the center of an emerging “agentic web” of autonomous systems interacting with one another. The releases spanned everything from a scientific discovery platform to a 

The announcement count continued apace as Google said it released 100 new tools at its own I/O conference. Most notable among these was a new AI Mode in search, a chat-search interface with multimodal and reasoning capabilities, and a Deep Search feature for more thorough research. (It wasn’t immediately well-received: A statement from the trade group News/Media Alliance 

 feature “the definition of theft” for displaying publisher content without links to sources.) Google’s realistic new video model, Veo 3, which now incorporates audio, also grabbed 

. And capabilities developed through experimental agent Project Mariner will allow AI Mode searchers to do things like book tickets or shop, according to Google.

While OpenAI and Microsoft are sometimes portrayed as eating into Google’s core business, the search giant’s firehose of AI news may serve as a reminder that the AI race is still neck and neck.

Chirag Dekate, VP analyst at Gartner, said Google’s announcements around search were “incredibly disruptive and understated.” While Google’s announcements focused largely on consumer products, they could have big implications for businesses, Dekate said. The new agentic search experience might mean not just a rethink of SEO but also 

 and how companies interact with businesses online more broadly.

“Google is aggressively redefining search from a list of links to an AI-powered, conversational, and agentic experience. This has profound implications for the enterprise,” Dekate told Tech Brew in an email. “Businesses must adapt to a world where search is increasingly about direct answers and task completion, not just information retrieval. This necessitates a shift in SEO strategies, content creation, and how businesses engage with customers online.”

Close up of a megaphone with pixels following an AI profile icon.

Tech giants one-up each other in a weeklong AI news blitz

From a new Google search to a devious Anthropic model, here are some highlights from a busy week in AI news.

Systems like ChatGPT have now been trained on the near-sum of all human knowledge. Yet they spend much of their time whipping up emails or regurgitating facts. What would it take for generative AI to spawn new ideas?

That’s a question that tech companies and research labs seem to be increasingly asking as they explore how GenAI might aid in the process of scientific discovery. New AI systems aim to give scientists tools to supercharge the hypothesis and experimentation flow.

OpenAI is readying the release of new reasoning models—o3 and o4-mini—as soon as this week that are said to be better at drawing from different scientific fields to put forth novel ideas and devise experiments, which could be helpful for things like designing new drugs or materials, 

. The company has been holding “AI jam sessions” at national labs to allow hundreds of scientists to explore how to put reasoning models and other AI offerings to use.

Google said at its Cloud Next conference last week that it would be making scientific tools like AlphaFold 3 and WeatherNext available through Google Cloud. The systems are based on Google DeepMind’s research on 

, respectively. Google previously released a multi-agent system called Co-Scientist in February, though some researchers 

A startup called Lila Sciences raised $200 million in seed capital last month on its promise to build a “scientific superintelligence platform” and “fully autonomous labs” across biology, chemistry, and materials science. The company claims its tech can help scientists formulate hypotheses and design experiments.

Many of these efforts are fueled by breakthroughs in AI reasoning that allow models to “think” through complex problems beyond just summarizing and amalgamating. GenAI has already been yielding progress in fields like 

—for which Google DeepMind won a Nobel Prize last year—

Yet some AI researchers also say that fully unlocking the potential of GenAI might require a rethink of how AI models are trained and tested. And some 

 on the lofty promises of AI company leaders in this arena.

GenAI’s role in scientific discovery has been steadily growing in recent years, Vanessa Parli, research director at Stanford’s Human-centered AI Institute, 

 last week. Parli said the institute added a chapter about scientific discovery to its annual AI Index for the first time last year and expanded it even further this year.

“AI’s role in scientific progress continues to expand,” the report says. “While 2022 and 2023 marked the early stages of AI-driven breakthroughs, 2024 brought even greater advancements.”