Signals Inbox·August 27, 2026·AI Infrastructure

What startups have Google bought recently?

Google has recently bought ProducerAI, Common Sense Machines, Atlantic Quantum, Galileo AI, Evaluable AI, MutableAI, AdHawk Microsystems, Cameyo, Wiz and Intersect, while the splashy Windsurf, Character.AI and Hume AI deals were not full acquisitions.

Q1Which startups has Google actually bought recently?

Google has bought at least ten clearly identifiable startups or private technology companies in its latest acquisition cycle, and the mix is much broader than AI software.

The most recent full startup acquisition we can verify is ProducerAI, the generative-music company that joined Google Labs. Before that came Common Sense Machines in 3D AI and Atlantic Quantum in quantum hardware. Google also acquired smaller companies including Galileo AI, Evaluable AI, MutableAI, AdHawk Microsystems and Cameyo.

Then there are the two deals that completely change the scale of the list. Google completed its $32 billion acquisition of cloud-security company Wiz, while Alphabet completed its acquisition of energy and data-center developer Intersect after agreeing to pay $4.75 billion in cash plus debt. Alphabet's latest financial filings put the adjusted purchase prices at about $29.5 billion for Wiz and $5.9 billion for Intersect.

Several deals that often get described as Google acquisitions do not belong on this list. Google never bought Windsurf, Character.AI or Hume AI outright. We will come back to those because they reveal a second acquisition strategy running alongside the conventional one.

Recent Google startup acquisitions

Q2Is Google buying startups as often as it used to?

Google is buying far fewer companies than during its old acquisition frenzy, even though the deals it cares about today can be dramatically larger.

The contrast with the early 2010s is huge. Google completed 48 acquisitions in 2010 and 79 in 2011, according to its filings. In 2011 alone, it was effectively closing six or seven acquisitions every month.

The total cost of those 79 acquisitions was roughly $1.9 billion. One recent Google acquisition can now cost many times that entire annual bill.

Deal count is a bad way to judge Google's current appetite. During the old buying spree, Google regularly picked up small products and teams. These days, it still makes small technical acquisitions, but it is also willing to make a very large bet when a company controls something Google considers strategically important.

The pace has slowed. The conviction behind the biggest deals has gone sharply in the opposite direction.

Q3Why is Google suddenly spending so much on acquisitions?

Google's recent acquisition spending is concentrated around two expensive problems: securing AI workloads and getting enough physical infrastructure to run them.

Wiz and Intersect together represented at least $36.75 billion of announced cash consideration before debt and purchase-price adjustments. Wiz accounts for roughly 87% of that amount.

That concentration is useful because it stops us from drawing the wrong conclusion from the long list of small AI acquisitions. Google has bought companies working on music, 3D generation, interface design and code understanding, but those startups are not where the disclosed acquisition money went.

The huge checks went toward cloud security and data-center capacity.

Google Cloud is also growing quickly enough to make those priorities understandable. Alphabet reported that Google Cloud revenue grew 36% in 2025, and the company said in a recent investor presentation that 75% of Cloud customers were already using its AI products. It also said enterprise AI had become Cloud's main growth driver.

Google now has a lot more at stake when companies move important AI workloads onto its infrastructure. Security, electricity and available data-center capacity can directly limit that growth.

Q4Why did Google pay $32 billion for Wiz?

Google's $32 billion Wiz acquisition looks expensive even by Big Tech standards, but Wiz had grown into one of the strongest independent cloud-security companies before Google bought it.

Wiz was founded in 2020 and reached roughly $100 million in annual recurring revenue unusually quickly. By 2025, reporting around the acquisition put Wiz above $1 billion in ARR.

Google had already tried to buy the company for roughly $23 billion in 2024. Wiz walked away and continued preparing for an IPO. Google returned the following year with a $32 billion agreement.

That means Google's offer rose by roughly $9 billion after Wiz rejected it. The extra money came as Wiz kept growing rather than after its position weakened.

The size also makes more sense next to Google's previous security purchases. Google paid about $5.4 billion in announced consideration for Mandiant in 2022. Wiz cost almost six times as much.

Google Cloud now gets a security platform that already works across AWS, Microsoft Azure and Google Cloud. That multicloud position is important because many large companies do not want a security product tied to only one cloud provider.

Wiz also remains a recognizable brand inside Google Cloud today. Google bought an existing platform, customers and distribution rather than simply absorbing a security engineering team.

Q5Why did Google buy an energy company like Intersect?

Alphabet bought Intersect because electricity and ready-to-use data-center capacity have become serious constraints on Google's AI expansion.

Intersect develops large energy projects alongside data-center infrastructure. Before the acquisition, Google had already invested in the company and was working with it on projects designed to bring new power and computing capacity online together.

The logic is fairly practical. Google can order more accelerators and build more AI models, but those systems still need buildings, grid connections and enormous amounts of electricity.

Alphabet's latest filings say the Intersect acquisition is specifically meant to accelerate data-center capacity and energy development. The accounting value came to about $5.9 billion after purchase-price adjustments.

Intersect said before the acquisition that it had developed more than 11.5 gigawatts of clean-energy capacity across 14 facilities, with more than $20 billion of infrastructure operating or under construction.

Google is therefore moving deeper into the physical side of computing. Data centers used to look like supporting infrastructure around the software business. AI has made the ability to secure power and bring capacity online quickly much more strategic.

Q6Which small AI startups is Google buying around Gemini?

Google is currently buying small AI companies that fill very specific gaps around Gemini, especially in evaluation, coding, interface generation and 3D content.

Evaluable AI is one of the clearest examples. The startup built tools for comparing LLM outputs and measuring model quality. Its founders later announced that Google had acquired the company.

MutableAI worked on AI coding tools before moving toward automatic codebase documentation. Its Auto Wiki product could turn a software repository into continuously updated explanations and diagrams. After the acquisition, founder Omar Shams said the team rebuilt that idea inside Google as CodeWiki.

Galileo AI attacked a completely different problem. It allowed people to generate user-interface designs from prompts. After Google acquired the company, co-founder Arnaud Benard announced that Galileo's work was continuing through Stitch, Google's AI interface and app-design product.

Common Sense Machines pushed further into generative 3D. The Cambridge startup developed models that could create 3D assets from ordinary images. Google confirmed the acquisition, and its roughly dozen-person team joined Google DeepMind. The company had reportedly last been valued at only around $15 million.

Put those four acquisitions together and a pattern appears. Google already spends enormous amounts building foundation models internally. What it keeps finding outside are small teams that have gone unusually deep on one useful layer around those models.

The acquisitions give Google shortcuts into problems that would otherwise need another internal team, another product cycle and another round of experimentation.

Q7Why buy ProducerAI when Google already has its own AI music models?

ProducerAI makes sense because a strong music model does not automatically produce a strong music-making product.

Google DeepMind already had Lyria when ProducerAI joined Google Labs. ProducerAI, previously known as Riffusion, had spent years building the interface around AI music creation.

That difference shows up in the product. ProducerAI lets people work on lyrics, melodies, instruments, effects and other parts of a song rather than simply asking an AI model to generate one finished track.

Google quickly connected the product to its own technology. ProducerAI started using Gemini, Lyria, Veo and Google's image-generation models, then became Google Flow Music.

The next move was commercial rather than experimental. Google partnered with music company Believe to put Flow Music and Lyria 3 Pro in front of artists distributed through Believe and TuneCore.

So the acquisition brought Google something its research lab could not get simply by improving Lyria's benchmark performance: a team that had already spent years figuring out how people might actually create music with generative AI.

Q8Is Google buying hardware startups too?

Google is buying hardware startups where it sees a chance to control technologies that could become important to quantum computing or future wearable devices.

AdHawk Microsystems is a good example. The Canadian startup developed eye-tracking systems that avoid conventional cameras and use low-power MEMS technology instead. Bloomberg originally reported a deal worth about $115 million, including performance payments. Later filings from one of AdHawk's investors confirmed that Google had acquired the company.

Eye tracking could become particularly useful if Google succeeds in bringing Android XR to lightweight smart glasses. Power consumption, size and latency matter much more on glasses than on a desktop headset.

Atlantic Quantum tackles a completely different hardware bottleneck. The MIT spinout built superconducting quantum hardware that integrates qubits and control electronics more tightly inside the extremely cold environment used by quantum computers.

Google Quantum AI said the technology should help it scale its own superconducting-qubit architecture. Atlantic Quantum had raised only about $13 million, according to CB Insights, so this was a small company attacking a highly specialized engineering problem.

Google has already proved with its Willow processor that it can build impressive quantum hardware internally. Atlantic Quantum gives that existing program another technical approach to one of its hardest problems: turning relatively small quantum systems into much larger error-corrected machines.

Q9Did Google actually buy Windsurf?

Google did not buy Windsurf, even though the $2.4 billion deal was widely described in language that made it sound like an acquisition.

Google obtained a non-exclusive license to Windsurf's technology and hired CEO Varun Mohan, co-founder Douglas Chen and a group of researchers. Windsurf itself remained a separate company.

The difference became obvious almost immediately. Cognition, the company behind Devin, subsequently acquired the remaining Windsurf business.

The sequence was unusual. OpenAI had previously been discussing an acquisition of Windsurf reportedly worth around $3 billion. That agreement expired. Google then struck its $2.4 billion licensing and hiring deal without purchasing the company.

For Google, the outcome was still valuable. It got important Windsurf technology and many of the people behind it without taking ownership of the entire startup.

For a list of "startups Google bought," though, Windsurf should be excluded.

Q10Did Google buy Hume AI or Character.AI?

Google did not buy Hume AI or Character.AI either, although both companies gave Google technology and important employees through large licensing and hiring agreements.

Character.AI was the bigger transaction. Alphabet disclosed that it paid roughly $2.7 billion for a non-exclusive license to Character.AI technology and hired founders Noam Shazeer and Daniel De Freitas along with other employees.

Character.AI continued operating independently.

Hume AI followed a similar route. Google DeepMind obtained rights to some of Hume's voice technology while CEO Alan Cowen and around seven engineers moved to Google. Hume continued under new leadership.

The Hume deal is particularly interesting now because voice has become much more important across AI assistants, wearables and multimodal models. Hume had spent years developing systems designed to understand emotional cues in speech, while Google is pushing Gemini into increasingly conversational interfaces.

Google licensing-and-hiring deals that were not full acquisitions

Q11Why does Google keep doing deals that stop short of buying the startup?

Google can get much of the value of an AI startup by hiring its best people and licensing its technology, while leaving the remaining company outside Google.

That structure works particularly well in AI because a startup's value can be heavily concentrated in a few researchers, founders and pieces of intellectual property.

Buying the whole company also means buying everything else attached to it: contracts, employees Google may not need, liabilities and a corporate entity that has to be integrated.

These arrangements have become controversial for exactly that reason. The FTC has said it is looking more closely at Big Tech acqui-hires, and competition authorities in other countries have also started examining licensing-and-hiring transactions.

We should therefore treat them as their own category. Calling them acquisitions exaggerates how many startups Google actually owns. Ignoring them would miss some of Google's most important startup deals.

Q12Did Google recently buy Intrinsic?

Google did not recently buy Intrinsic from an outside owner because Alphabet already controlled the robotics company.

Intrinsic began inside Alphabet's X research lab and became a separate Alphabet company in 2021. It builds software and AI systems for industrial robotics.

What changed recently was the corporate structure. Alphabet moved Intrinsic into Google so the robotics team could work more closely with Gemini, Google DeepMind and Google Cloud.

That is strategically relevant even though no outside acquisition happened. Google is currently trying to connect its AI models more directly to physical machines, and bringing Intrinsic closer to its core AI teams makes that easier.

Some acquisition databases count the move as an acquisition because Google became the direct owner. From Alphabet's perspective, the asset was already inside the group.

We would therefore leave Intrinsic out of a clean list of startups Alphabet recently purchased from external shareholders.

Q13Is Google mainly buying AI companies now?

Google's small acquisitions are heavily tilted toward AI today, while its biggest purchases show that Google is equally worried about everything needed to deploy AI at scale.

Look at the smaller companies first. Evaluable AI worked on model evaluation. MutableAI worked on code understanding. Galileo AI generated interfaces. Common Sense Machines generated 3D content. ProducerAI built AI music tools.

Then the stack moves into hardware. Atlantic Quantum contributes quantum-computing technology, while AdHawk brings eye tracking for possible wearable interfaces.

The expensive deals sit further down the infrastructure stack. Cloud security protects workloads, while energy and data-center development determines how much computing capacity Google can actually deploy.

Meanwhile, the type of acquisition that used to change Google's consumer footprint has become much rarer. YouTube brought Google video distribution. Android gave it a mobile operating system. Fitbit gave it millions of wearable users. Google's recent purchases mostly bring technology, infrastructure or specialist teams.

Look at the portfolio vertically and it makes more sense. Google is filling gaps between electricity, computing infrastructure, models and the products people ultimately use.

Where Google's recent acquisition activity sits in the stack

Q14Is Google getting more aggressive with startup acquisitions?

Google is more aggressive with acquisition dollars today, but much more selective about where it spends them.

The early-2010s version of Google bought companies constantly. The current version appears willing to let dozens of startups remain independent while putting huge amounts of money behind a few areas that management sees as strategically urgent.

The smaller purchases reinforce that strategy. Google continues to pick up tiny specialist teams when they have technology that fits directly into Gemini, Google Labs, Android XR or Google Quantum AI.

It also has another route when ownership is unnecessary. Licensing-and-hiring deals let Google pull important talent and technology from startups without acquiring the full company.

In practice, Google now has three distinct ways of dealing with startups: full strategic acquisitions, small technical acquisitions and reverse acqui-hires.

The number of conventional deals alone therefore understates how active Google has become around startup technology.

Q15So, what startups has Google bought recently?

Google has recently bought companies including ProducerAI, Common Sense Machines, Atlantic Quantum, Galileo AI, AdHawk Microsystems, Evaluable AI, MutableAI, Cameyo, Wiz and Intersect, while several famous AI deals such as Windsurf, Character.AI and Hume AI were not full acquisitions.

Google's current buying strategy is unusually focused.

The company is still acquiring small teams when they have a useful piece of AI, quantum or wearable technology. At the same time, it has shown that it will spend tens of billions when management believes an infrastructure layer could become a bottleneck.

That combination is more revealing than the raw deal count. Google currently seems interested in owning or controlling whatever can make its AI stack stronger, faster to deploy or harder to disrupt, from power and cloud security all the way up to coding, design, 3D and music products.