Signals Inbox·July 28, 2026·Frontier AI

Who is the next Google?

OpenAI is the closest candidate to become the next Google because ChatGPT has become a starting point for complex digital work, but Google still owns the distribution, cash flow and ecosystem that make the comparison premature.

Q1What would a company actually have to do to become the next Google?

A company becomes the next Google only when millions of people start there by default and an entire business ecosystem grows around that habit.

Google’s original breakthrough was search, yet search alone does not explain what Google became. The company spent years connecting a daily consumer habit to advertisers, websites, browsers, phones, developers, cloud customers and its own computing infrastructure. A popular AI assistant can challenge Google’s first point of contact without reproducing the rest of that machine.

For this article, “the next Google” means the company most likely to become the main starting point for finding information and completing digital tasks, then turn that position into a durable platform. We are looking for habit, money, distribution and ecosystem control. Model benchmarks help, but they move too quickly to settle the question.

The four tests for becoming the next Google

Q2Why are people asking whether OpenAI is the next Google now?

People are asking because ChatGPT has already created a global consumer habit at a speed no previous standalone app matched.

OpenAI currently reports more than 900 million weekly ChatGPT users and over 50 million paying consumer subscribers. Sensor Tower separately found that ChatGPT became the fastest mobile app to reach one billion monthly users. Across generative AI apps, time spent during the first half of this year is projected to reach 36 billion hours, more than double the same period a year earlier.

The usage itself tells us more than the headline totals. People use ChatGPT for writing, research, coding, planning, learning and increasingly for searches that would previously have started with Google. OpenAI says search activity inside ChatGPT has almost tripled in a year. The product occupies the beginning of many tasks, the position Google spent two decades defending.

The shift is concentrated in certain tasks. People open ChatGPT deliberately when conversation or creation helps. Google remains woven into navigation, local queries, shopping, news, websites and countless quick lookups. ChatGPT has built the first credible new doorway to online knowledge. Google still owns the much larger building around it.

Q3Is Google Search actually losing to ChatGPT today?

Google Search remains commercially dominant today, even as ChatGPT takes a growing share of complex questions and knowledge work.

Alphabet’s latest quarter makes the position unusually clear. Google Search and related revenue exceeded $63 billion, up 17% from a year earlier. AI Mode passed one billion monthly users, while the Gemini app reached 950 million monthly users and tripled its daily audience over the year. Google says the addition of AI features is increasing total search queries rather than shrinking them.

Independent traffic estimates tell a similar story. Google’s conventional search share remains around 90% worldwide, and recent Similarweb and Sensor Tower estimates showed Google web visits and mobile usage still rising while AI assistants grew much faster from smaller bases. The market is changing, but Google’s core business has yet to fall.

Google has lost ground in one valuable area. ChatGPT now captures part of the moment when someone wants an explanation, comparison, draft, analysis or plan. Google is answering by putting Gemini directly inside Search, YouTube, Workspace and Cloud. So far, that response is protecting revenue while building a large AI audience.

Google Search and Gemini at their latest reported levels

Q4Has ChatGPT become the new place people start online?

ChatGPT is now the first stop for many writing, research, coding and planning tasks, although it still covers a narrower slice of daily online behavior than Google.

ChatGPT’s lead over other dedicated AI assistants is wide. OpenAI says the product receives six times as many combined web visits and mobile sessions as the next-largest AI app. It also reports four times more total time spent than that rival and more time than all other AI apps combined. Those comparisons come from OpenAI, but Sensor Tower’s separate finding that ChatGPT reached one billion monthly mobile users supports the broader conclusion.

Even with that lead, ChatGPT remains far from becoming the internet’s homepage. Most people still use browsers, search engines, social apps, maps, marketplaces and messaging services throughout the day. ChatGPT wins when a task requires synthesis or production. Google wins far more of the small, frequent actions that make up everyday internet use.

For now, ChatGPT is the default general AI assistant for a very large audience. Reaching the broader role once held almost entirely by Google would require deeper control over navigation, commerce, local information, devices and third-party services.

Q5Can OpenAI turn ChatGPT into a Google-sized money machine?

OpenAI can build a huge business around ChatGPT, but it has yet to prove the margins and marketplace power that made Google exceptional.

OpenAI says it is generating about $2 billion in revenue each month, equivalent to a $24 billion annual run rate. The company reached that level remarkably quickly, and more than 40% of its revenue comes from enterprise customers. Its early advertising pilot also crossed a $100 million annual run rate within six weeks.

The comparison with Google is still brutal. Alphabet’s latest quarter produced $119.8 billion in revenue and $40.8 billion in operating income. Google can finance models, chips, data centers and distribution from businesses that already throw off enormous cash. OpenAI is building its revenue engine while paying for rapid growth in computing demand.

Subscriptions and enterprise contracts may eventually support excellent margins as the cost of running AI requests falls. Advertising and shopping could add another layer, especially when users express clear intent. Conversational advertising also creates a harder trust problem than sponsored links on a results page. OpenAI needs to show that commercial recommendations can generate revenue without making answers feel purchased.

OpenAI will probably end up with a hybrid business: subscriptions for heavy users, usage-based enterprise revenue, developer fees, advertising for free users and transaction income from commerce. That mix could produce one of the world’s largest companies. Matching Google’s economics still requires those streams to reinforce one another while the cost of serving each task falls sharply.

Q6Is OpenAI becoming a platform companies cannot easily leave?

OpenAI is embedding ChatGPT deeply enough in business workflows that leaving it is starting to become expensive.

The latest growth is moving beyond individual chat sessions. OpenAI reports more than two million weekly Codex users, up fivefold in three months, while enterprise products now account for over 40% of revenue. Its APIs process more than 15 billion tokens per minute. Companies are using the same provider for employee assistance, software development, customer service, internal search and custom agents.

A company behaves differently once it connects internal data, writes evaluations, builds agents, trains employees and redesigns processes around one platform. Changing provider then becomes expensive. Individual models can still be swapped for some tasks, but the surrounding workflow has weight.

OpenAI’s weak spot is ownership of the surrounding environment. Microsoft, Google, Amazon and Oracle provide much of the cloud capacity. Apple and Google control mobile operating systems. Browsers, office software and enterprise identity systems also belong largely to incumbents.

Without a phone or cloud of its own, OpenAI must keep enough direct user demand and workflow dependence to stop its suppliers from becoming gatekeepers.

Q7Is Anthropic’s Claude becoming the default AI at work?

Anthropic’s Claude has become OpenAI’s strongest challenger for paid workplace use, especially coding and long-running enterprise tasks.

Anthropic says its annual revenue run rate has crossed $47 billion. Claude Code alone exceeds a $2.5 billion run rate, more than double its level at the beginning of the year. The company also reports that over 1,000 customers spend at least $1 million a year with Anthropic. These company disclosures place Claude far beyond a niche assistant for programmers.

Claude makes its best money from work companies value highly: coding, document analysis, research and agents that run for long periods. Anthropic also created the Model Context Protocol, an open way for assistants to connect with business tools and data. Its official Python and TypeScript software kits reached more than 97 million monthly downloads before Anthropic donated the protocol to the Linux Foundation.

Claude still trails badly in consumer habit. A recent survey of 1,999 U.S. AI-assistant users found that 7% primarily used Claude, compared with 58% for ChatGPT and 25% for Gemini. Anthropic also distributes Claude through AWS, Google Cloud and Microsoft Azure, gaining enormous reach while leaving part of the customer relationship in other companies’ hands.

Claude’s clearest future is as the default system for serious knowledge work. That narrower position can become extraordinarily valuable if companies use several consumer assistants but standardize professional agents around Claude and MCP.

Q8Can Meta make its AI unavoidable through WhatsApp, Instagram and Facebook?

Meta can place an AI assistant in front of more people than any startup. Social distribution alone will not make Meta AI the place where people begin serious tasks.

Meta’s family of apps reaches 3.56 billion people each day. The company generated $56.3 billion in its latest reported quarter, with advertising impressions up 19% and the average price per ad up 12%. It can place AI inside WhatsApp, Instagram, Facebook and Messenger while funding the rollout through an advertising business that is still expanding quickly.

Lately, Meta has pushed its assistant beyond image generation and casual questions. Meta AI can now plan multi-step activities and follow through on tasks, while Facebook has added an AI search mode built around public recommendations and discussions from Meta’s apps. Millions of people also use Meta’s AI glasses each day, giving the company an early route into hands-free computing.

The harder problem is getting people to choose Meta AI for serious work. People open WhatsApp to message someone and Instagram to watch or share content. An assistant sitting inside those products can gain huge exposure without becoming the tool users trust for research, coding, business decisions or complex planning.

Meta can still win a huge part of AI by making the assistant ambient across communication, entertainment, advertising and wearable devices. That would give the company enormous influence over everyday use, even if ChatGPT, Gemini or Claude remains the chosen tool for harder work.

Q9Could Microsoft win AI because companies already live inside its software?

Microsoft has the easiest route to making money from AI inside companies because its customers already pay for Azure, Microsoft 365, GitHub and security tools.

Microsoft’s latest reported quarter produced $82.9 billion in revenue, up 18%, and $38.4 billion in operating income, up 20%. The company is spending heavily on data centers and chips, but every new capability can be sold through products companies already use to write documents, manage email, build software, protect systems and run cloud workloads.

For buyers, the choice can be simple. A company considering an AI assistant may prefer the product already covered by its Microsoft contract, identity controls, security policies and data permissions. GitHub gives Microsoft a similar advantage with developers. Azure earns from model demand even when another company supplies the model itself.

Microsoft’s consumer position is less clear. Copilot refers to several products across Windows, Microsoft 365, GitHub and the web, which makes the brand harder to understand than ChatGPT. The company can win a huge share of AI spending without becoming the place ordinary people instinctively begin a question.

Q10Could Apple decide the next Google through the iPhone and Siri?

Apple can steer billions of AI interactions through iPhone defaults, even if another company supplies some of the underlying intelligence.

Apple’s installed base exceeds 2.5 billion active devices. Its newly introduced Siri AI is designed to use personal context, understand what appears on screen, search broad world knowledge and take actions across apps. A capable assistant delivered through the operating system reaches users before they choose ChatGPT, Gemini or Claude.

An iPhone already has access to identity, payments, location, messages, photos and apps. Those permissions could make Siri more useful for personal tasks than a separate chatbot, provided Apple executes well. The company’s earlier delays in advanced AI leave that outcome uncertain, and the new system has only recently been introduced.

Apple’s power comes from controlling access. It can favor its own models, call outside providers for selected tasks or let users choose among several assistants. Each option changes who receives the query, who keeps the context and who owns the customer relationship.

Apple’s choices may determine how fragmented the market becomes. A strong Siri could prevent any independent assistant from gaining Google-like control on mobile. A weak Siri could turn the iPhone into the most valuable distribution channel for whichever outside provider Apple places closest to the user.

Q11Does the best AI model automatically become the next Google?

The best AI model changes too often to determine the winner by itself.

Model rankings change with almost every major release from OpenAI, Anthropic, Google or xAI. Customers can also reach several models through one cloud platform, coding tool or routing service. A technical lead measured in months rarely creates the decade-long protection Google built around search.

User habits and integrations move much more slowly. Google can introduce a new Gemini model to Search, Android, Workspace, YouTube and Cloud. Microsoft can add a model to Office, GitHub and Azure. Meta can place one across its social apps. OpenAI has the strongest independent consumer habit, while Anthropic has gained a deep position in professional work.

The same pattern already appears in coding. Cursor has built a large business while using and evaluating models from several providers. Companies often care about the complete tool—context management, reliability, permissions and user experience—more than the name of the model handling every request.

A top-tier model is required. Once several providers reach that level, product design, cost, trust, distribution and existing workflows decide the winner.

Q12Is NVIDIA capturing more value than the companies building AI assistants?

NVIDIA currently gets paid by nearly every serious AI competitor, giving it the safest business model in the race even without owning the assistant people use directly.

NVIDIA’s latest quarter produced $81.6 billion in revenue, up 85% from a year earlier. Data Center revenue reached $75.2 billion, up 92%. The industry is pouring money into chips and computing systems before many AI apps have proved they can earn lasting profits.

NVIDIA also benefits from CUDA, networking equipment and complete systems that developers and data-center operators have spent years learning to use. The advantage extends beyond one generation of chips. It includes software, engineering habits and a supply chain designed around NVIDIA hardware.

NVIDIA rarely owns the main relationship with the person asking a question or the company automating a process. Its customers can try custom chips from Google, Amazon and others as the market grows. Alphabet alone spent $44.9 billion on capital projects in its latest quarter, mostly for AI infrastructure, while OpenAI and Anthropic are spreading workloads across several chip and cloud partners.

NVIDIA may capture more profit from the AI build-out than any single assistant company. It is the essential computing supplier of the era, which is a very different job from becoming the service people use to navigate information and complete tasks.

Q13Can Perplexity, xAI or Cursor still become the next Google?

Perplexity, xAI and Cursor can still build dominant businesses, but each owns a narrow doorway rather than a general-purpose starting point.

Cursor has the clearest proof of strong economics. The company passed $1 billion in annualized revenue and says more than 70% of the Fortune 500 use its coding agents. Its product can become the default place where software teams plan, write, review and maintain code, even while Cursor relies on several underlying model providers.

xAI has a different asset: immediate distribution through X, Grok apps and Tesla vehicles. The company describes a combined reach of about 600 million monthly users across X and Grok, although that reach includes the much larger X audience. Grok also benefits from real-time material on X and has recently expanded its office and coding capabilities.

Perplexity remains the most focused independent attempt to rebuild search around direct, sourced answers. Its product is easy to understand and can win users who want research instead of an open-ended chatbot. Scale is the problem. Google and OpenAI can copy many search features while distributing them to much larger existing audiences.

A specialist can expand from one powerful habit, just as Google expanded from web search. None of these companies has the breadth of consumer behavior, distribution, infrastructure and monetization needed for the full title today.

Q14Will AI answers damage the web enough to slow the next Google?

AI answers are already taking clicks from publishers, and that could weaken the information supply every assistant depends on.

A large study of more than 55,000 Google queries found that AI Overviews appeared for 13.7% of queries overall and 64.7% of questions. Researchers broke the generated answers into 98,020 factual claims and found that 11% lacked support from the cited pages. Another study using Wikipedia’s multilingual rollout estimated that exposure to Google AI Overviews reduced daily visits to affected English-language articles by about 15%.

Lost traffic is only part of the problem. Research covering 24,000 queries across 243 countries found that AI search presented fewer small specialist sources and less varied information than conventional search. A separate analysis of more than 200,000 real ChatGPT interactions found that AI allowed people to ask a broader range of questions, yet answers to comparable searchable questions were usually less diverse than Google results.

Publishers can respond with licensing deals, subscriptions, paywalls and restrictions on crawlers. Wikipedia has already signed commercial data agreements with several AI companies. Larger publishers have similar bargaining power. Small specialist sites often do not.

The next Google will need a healthier exchange with information producers than taking the material while removing the click. Otherwise, fewer independent sources will be able to fund reporting, testing and expertise. The assistant becomes more convenient while the knowledge underneath it gets thinner.

Q15What does OpenAI still need to prove before it deserves the next Google label?

OpenAI still needs its own distribution, strong margins and customer workflows that are hard to move before the next Google label becomes convincing.

OpenAI has proved that people will build a daily habit around ChatGPT. It also has a credible enterprise business, a fast-growing coding product and an early advertising operation. What we still do not know is whether the lead will last and pay. Can OpenAI hold its audience as Gemini and Claude improve? Can it serve agents cheaply enough to produce high margins? Can it own customer relationships while relying on outside clouds, chips, browsers and mobile operating systems?

Commerce could become especially important. Google built an auction around commercial intent. OpenAI needs an equally natural way to earn when a user researches a purchase, books a trip, selects software or asks an agent to complete a transaction. The ads pilot is growing quickly, but six weeks of revenue does not establish a mature marketplace.

The label becomes convincing only after OpenAI shows several years of falling cost per task, strong free cash flow, developers who cannot easily move and direct distribution that incumbents cannot easily restrict. Its trajectory makes each outcome plausible. User growth alone guarantees none of them.

What OpenAI must prove to become the next Google

Q16So who is the next Google?

OpenAI is currently the closest candidate, but no company has earned the title yet.

ChatGPT has achieved the part that once looked least likely: hundreds of millions of people deliberately open a new service to ask questions, create work and make decisions. OpenAI has also converted that habit into substantial consumer, enterprise and developer revenue. Among companies founded in the AI era, it stands clearly ahead.

Google remains the stronger company and the more complete platform today. Its search revenue is growing, Gemini has reached mass-market scale, Android and Chrome protect distribution, and Alphabet can fund infrastructure from a profitable advertising and cloud system. Google is absorbing the new interface faster than challengers are dismantling the old business.

Anthropic could own a large part of professional intelligence. Microsoft may collect the most enterprise spending. Meta can make AI present across communication and social life. Apple can influence which assistant reaches mobile users first. NVIDIA will earn from almost everyone building the underlying systems.

OpenAI looks most like the next Google when we focus on where people begin complex digital tasks. It falls well short when the comparison includes distribution, cash generation, infrastructure and control of a wider ecosystem.

The market is more likely to split among several powerful layers than hand everything to one replacement. OpenAI leads the new consumer interface. Google still owns the broadest system around it. The title belongs to OpenAI only if ChatGPT keeps its default status long enough to build that missing system.