Google I/O: Alphabet’s AI showcase has a chance to surprise Wall Street

alphabet‘s stock is up 140% in the past year, and the cloud business is Amazon’s and microsoft’s.

But 18 months ago, it looked like Google’s parent company had spent a decade preparing for the age of artificial intelligence and was just watching OpenAI define the market.

Wall Street now views Alphabet as one of the few companies that stands to benefit from all the tiers of the generative AI boom.

Google I/O, which begins Tuesday, has always been a place to show developers where the company is headed. This year, the stakes are even higher.

Wall Street is already praising Alphabet’s AI resurgence, but investors want to know whether that confidence is backed up by an actual product roadmap across key areas such as search, cloud, Android, chips and enterprise software.

“Google is probably the company in the best position to monetize AI at scale because it controls almost every layer of the stack,” said Lo Tony, founding managing partner at Plexo Capital and an early investor in Anthropic. “I haven’t really seen a company that has complete vertical integration from top to bottom to support AI.”

Gene Munster, managing partner at Deepwater Asset Management, said the advantage of having so many layers of control is not just scale, but speed.

“Owning the full stack has advantages in terms of speed of innovation,” Munster said. “For example, if you’re building on your own custom silicon, this is a speed advantage. If you have access to power, you can get your data center up and running faster. That’s a speed advantage, and that’s important.”

Here are seven key areas investors are watching for Google I/O.

The most interesting announcement will be whether Google will announce the next generation Gemini model.

Reports ahead of I/O point to a possible Gemini 4 debut, but analysts aren’t entirely counting on it. Citi noted that with Gemini 3.1 Pro released in February, Google is releasing at a roughly 3-4 month pace, making a Gemini 3.2 or 3.5 update more likely than a full generational leap.

So questions about Gemini 4 go beyond the version number. Stepping up could give Google a clearer answer to OpenAI and Anthropic. Mizuho wrote that Gemini 4’s announcement “will push Google back to the forefront of the frontier,” but with just another third-generation update, it will be more of a catch-up.

Updates to the broader Gemini ecosystem will also be key.

Mizuho analysts said they are particularly focused on progress with Project Astra, Google’s universal AI assistant, as well as Gemini Live’s deeper functionality and use of native tools across screen sharing, video understanding, search, Gmail, Calendar and Maps. Updates to Google’s open source model families Gemma and Gemini Robotics are also coming.

Usage for the event is already higher than a year ago. Paid Gemini Enterprise monthly active users grew 40% sequentially in Q1. The Gemini app saw a 127% year-over-year increase in monthly active users in the U.S. in April, according to Citi data. As of Google Cloud Next, token consumption is reaching 16 billion per minute.

If there’s one theme that runs through the I/O Session lineup, it’s agents.

Google hosts sessions on agent coding workflows, multimodal tools, media generation, robotics, and AI agents. The goal is to position Gemini not just as a chatbot, but as an operating layer across Google’s products that can understand context and take action.

“It’s about who wins the office co-pilot market,” Tony said. “As larger markets become AI agents and orchestrate things like inference infrastructure, multimodal workflows, and enterprise search, we see a huge opportunity for Google to drive Alphabet’s future growth.”

Agenttic coding is part of positioning Gemini as a response to Anthropic’s Claude Code and OpenAI’s Codex. This category has become one of the clearest proofs of AI’s commercial value, especially in enterprise software.

Commerce may be a bigger opportunity. Google already has search, shopping, autofill, and payments. Now we want Gemini to be able to connect them to the agent checkout experience.

Google is expanding its Universal Commerce Protocol with additional partners, including: metaMicrosoft, Stripe, Klarna and affirm In the last few weeks. I/O is expected to further demonstrate how its infrastructure enables end-to-end agent checkout, meaning Gemini not only answers shopping queries but also completes transactions.

Sameer Samat, Google’s president of the Android ecosystem, explained that he asked Gemini to plan a barbecue, create a menu, open Instacart, add ingredients to the Safeway cart, and notify him when the task was completed.

“If you add this up multiple times a day over the course of a week, you’re looking back a considerable amount of time,” Sammat said. “I think these features will get people more excited and make it more tangible.”

Tony said Google’s multimodal experience provides a structural advantage as workflows become more complex.

“Enterprise workflows increasingly include video, audio, images, code, and more,” he said. “Google has a unique advantage with multimodal systems because they have this experience with some of the largest applications that handle multimodal systems, such as YouTube, Android, Maps, Search, DeepMind, and obviously with TPUs.”

For investors, the push for agent commerce has implications beyond Alphabet. Mizuho warned that more agent-like product development by Google could weigh heavily on markets such as: Reservation held, expedia, door dash, Jiro and instacartnoted that expectations for change are likely already contributing to some of the recent weakness in these stocks.

The next question is how Google will get paid for it. According to Citi, AI-powered campaigns now account for more than 30% of search spend. AI Max, which debuted in beta in April and is expected to replace dynamic search ads by September, is showing early results, including boosting conversions with a full suite of features.

Citi noted that AI mode could help Google monetize long, complex queries that have traditionally been difficult to turn into ad dollars. However, Mizuho points out the trade-off. The company estimates that AI mode searches see a significant drop in external clicks, with 93% ending without an external click, and AI summary queries see a 15% drop in organic click-through rates.

Therefore, monetization is one of the biggest challenges for I/O. Munster said he will focus on new advertising products within the AI ​​mode, how Google puts together agent commerce, and what it says about more personalized AI experiences.

But the most important I/O announcements for investors are likely to come from cloud and infrastructure.

Cloud has become one of Alphabet’s strongest pillars. It grew 63% year over year in Q1, outpacing both Azure and AWS. Cloud backlog is expected to reach $462 billion, up approximately 90% sequentially, with half of that amount expected to be recognized over the next 24 months. Gen AI product revenue increased approximately 800% year over year.

CEO Sundar Pichai cited several factors behind that growth, including faster new customer acquisition, greater commitment, and deeper relationships with existing customers.

On the company’s first-quarter earnings call, he said the number of deals worth $1 billion or more closed in 2025 is more than the previous three years combined, and that existing customers have exceeded original commitments by more than 30%.

The new wildcard is external TPU sales. Google revealed in the first quarter that it will begin offering custom AI chips to external customers in the second half of 2026, with plans for broader expansion in 2027. This is a potentially huge new revenue stream, but investors don’t yet know how to model it accurately.

Mizuho wrote that investors will want to hear details about whether external TPU sales are recorded as gross sales or royalty income, what the margins will be, and how those transactions will be recorded as backlog.

Toney said the TPU is one of the most undervalued parts of Alphabet’s thesis, and argued that Google’s in-house chips will allow the company to build a tightly integrated AI infrastructure that supports not only Gemini and Cloud, but also YouTube, Android, and other ecosystems.

Munster estimates the broader AI chip market is moving at about $500 billion a year, meaning even a small share increase could be a significant factor for Alphabet.

No relationship attracts more scrutiny of I/O than the relationship between Google and Anthropic.

Alphabet holds significant ownership stakes in AI startups, and its recently reported $200 billion cloud commitment, if accurate, could represent a large portion of Google’s contracted future cloud revenue.

Google has also committed to up to $40 billion in total investment, creating a loop where capital flows to Anthropic through cloud and TPU spending and back to Google.

This dynamic raises concentration issues that investors are already seeing elsewhere in the cloud. oracleThe company’s stock price soared after it reported a significant increase in its backlog, mostly related to OpenAI, but then sold off as investors grew concerned about customer concentration. Microsoft faces similar debate over its relationship with OpenAI.

Toney said Google’s relationship with humanity looks more like a hedge than a weakness.

He said that even if companies choose Claude over Gemini, Google could still benefit from the infrastructure demands behind its use.

“Even if companies prefer Claude, Google still wins in the infrastructure space, because all that activity has to exist somewhere,” he said. “Google still wins with TPU.”

WATCH: Inside the return on Alphabet’s capital investment: What investors need to know