In August 2023, as the AI craze was settling in for a sustained surge, The New York Times did a piece on Nvidia. Naveen Rao, then the head of software start-up MosaicML (he joined Databricks in August 2024 after it acquired the company), discussed Nvidia’s dominance and its implications for AI researchers: “Everybody builds on Nvidia first,” Mr. Rao said. “If you come out with a new piece of hardware, you’re racing to catch up.” This is reminiscent of the mid-1990s browser wars: web developers have traditionally built web sites with the primary goal of compatibility with the dominant browser of the day—first Microsoft’s Internet Explorer and later, Google Chrome.
The comparison is apt because the browser wars of the early 1990s never went away, and now they’ve become intertwined with the AI race. On July 9, Perplexity introduced its Comet browser to paying subscribers, and around the same time, news outlets reported that Open AI also plans to release its own browser, internally codenamed Aura. That means that along with Google’s Chrome, three of the top AI giants currently have browsers on offer. (Also, Microsoft has slowly been integrating its Copilot AI offerings into its Edge browser since September 2023.)
Why Do We Still Care About Browsers?
AI companies see browsers as integral to the competition to make their respective generative AIs increasingly useful to companies and individual users. It’s all well and good for generative-AI websites to Ghibli-ize family photos or help students write term papers, but the real reason investors have spent years pouring money into AI investments is that they hope the technology will spawn groundbreaking real-world applications (and, hopefully, real-world profits). The new browsers represent part of companies’ attempts to start delivering on that potential, though the sky-high capex and talent-recruitment spending by the major AI companies shows no signs of slowing.
Silicon Valley’s AI giants view agents as key to those efforts. Browsers, prosaic as they might seem some 32 years after the first popular browser (NCSA Mosaic) was introduced, are believed to be the way to build user stickiness to a given AI agent.
Agents are AI-powered software capable of figuring out how to complete a complex task with minimal instruction, navigating through third-party sites and applications when warranted to make decisions and take action. In marketing its Comet browser, Perplexity touted an example in which the user identified a dinner menu (butter chicken and chicken cobb salad), at which point the agent researched recipes from multiple websites, identified the ingredients commonly used in those dishes, and then populated an Instacart shopping basket with those ingredients. OpenAI’s Operator agent, introduced in January 2025, touts similar user-level abilities: Operator purportedly can assess numerous providers to find travel bookings that meet the user’s preferences and requirements, then make the chosen reservation automatically, filling out forms wherever needed.
It’s possible that someday, most of us will happily (or at least begrudgingly) pay a monthly fee for our own personalized AI agents to streamline our lives, armed with an intimate knowledge of our personal preferences, quirks, histories, and habits.
Beyond Consumers
The public sector is already using AI. In southern California, for example, authorities are hoping that AI bots can monitor thousands of cameras to provide early warning signs of wildfires before they become catastrophic. Some local U.S. police departments are using AI-powered bots for surveillance, deploying them on social media to masquerade as and interact with protesters and political activists.
But AI companies are also banking on a payoff from businesses. Companies and institutions in a broad range of fields and industries are already experimenting with artificial intelligence. For instance, we wrote recently about AI applications in health care, with providers putting AI to work in diagnostics, hospital management, and follow-up care.
To offer a few other examples, the biotech/pharmaceutical industries are also using AI to speed the development of new treatments and medicines. Some of the smaller upstarts in this space include Recursion Pharmaceuticals (RXRX N/A% ), which hopes its Recursion OS AI platform can combine analyses from internal and external datasets about human biology to dramatically speed every stage of the drug-discovery and development process. Meanwhile, AbCellera (ABCL -1.42% ) is hoping to leverage AI to identify novel antibodies. It has partnerships in place with larger pharmaceutical companies like Moderna and Eli Lilly.
Farmers have high hopes for AI as well. John Deere (DE -0.53% ) is developing autonomous farms that require minimal human tending, with a range of AI-driven machines like tractors, drones, irrigators, and harvesters using soil and weather data to take care of most tasks. Farmers appear especially hopeful that AI technology from companies like Deere and privately held companies like Carbon Robotics can provide an autonomous (and perhaps less chemically intensive) solution to identifying and eliminating weeds on a large scale. They’re also hopeful that AI weather-forecasting research being done at Nvidia, Google, and Microsoft might lead to more accurate predictions, for obvious reasons.
Major retailers like Walmart (WMT -1.21% ), Sephora, and of course, Amazon have all enthusiastically embraced AI as well, using the technology to forecast demand, manage inventories and supply chains, and make personalized customer recommendations. AI is also being broadly deployed in manufacturing (quality control, automation, predictive maintenance, etc.) and financial services (fraud detection and prevention, algorithmic trading, compliance, etc.) (We will save an exploration of AI within these sectors for another day.)
Still, as with individual users, the major AI companies appear to be particularly interested in persuading enterprises to pay for customized agents that can take over many of the day-to-day responsibilities of human employees—reading sales-tracking spreadsheets to generate and send out invoices; screening business leads; or handling customer inquiries.
Salesforce (CRM 0.13% ) has partnered with companies such as IBM (IBM -1.77% ), OpenAI, and Microsoft (Azure) to develop its own AI offerings, branded as Einstein. Einstein-based agents can help Salesforce clients by handling a full array of tasks in areas ranging from sales and marketing to analytics, customer service, human resources, and recordkeeping. Moving beyond single agents, Salesforce is also seeking to advance the growing agent-to-agent (A2A) trend, in which multiple agents created by enterprise users can interact with each other and with agents created by third parties or using competitors’ models.
Among Einstein’s major competitors are Accenture (ACN -0.95% ), which has built a growing book of business helping major enterprises including Unilever, ESPN, Best Buy, and BMW to deploy AI agents. Its AI Refinery platform, which helps businesses leverage Nvidia’s reasoning models to create and adapt AI agents with minimal technical expertise. Accenture also recently introduced its Trusted Agent Huddle, which enables multi-agent collaboration—both in-house and with agents created by third parties.
Microsoft (MSFT 0.26% ) has steadily been integrating its Copilot with cloud services, Edge browser, and business software such as its Office products for several years now. On an enterprise level, Microsoft offers its Azure AI Foundry platform, which is compatible with models from OpenAI, China’s DeepSeek, France’s Mistral, and others. The platform features templates that make it easier for business clients to produce and deploy their own agents that are compatible with open-source A2A protocols.
Google and ServiceNow (NOW 1.80% ) are also aggressively competing to help businesses create AI multi-agent platforms.
Taking Over the World
The AI industry isn’t stopping there. In the industry’s view, everyone needs an AI — even countries.
The British government seems to agree. On July 21, it signed an agreement with OpenAI to explore the use of its models in various government functions, including justice, defense, security, and education. The agreement is similar to the one it previously signed with Google DeepMind
Google’s agreement, for which the company agreed not to charge the British government, drew suspicion and alarm from British privacy advocates, some of whom questioned the wisdom of putting Britons’ private data in the hands of a U.S. company. Martha Dark, the co-executive director of Foxglove, a non-profit group that advocates on issues related to technology fairness, asked the Guardian, “How is the government going to be able to hold Trump-supporting U.S. big tech giants to any kind of serious account on this – or any other issue – after we’ve given Google the keys to the data kingdom? It’s hard to see this as anything other than dangerously naive.”
Setting aside Dark’s anti-Trump bias, it’s true that a number of countries are becoming more uncomfortable with their dependence on foreign companies for AI capabilities. This is driving a trend toward sovereign AI, in which AI infrastructure and personnel are located within a country’s own borders and in many cases, trained on the country’s own content. Nvidia’s Jensen Huang is in full agreement. As he argued, countries should develop sovereign AIs because this “codifies your culture, your society’s intelligence, your common sense, your history – you own your own data.” He further asserts that it doesn’t make sense for people to “export their country’s knowledge, their country’s culture, for somebody else to then resell AI back to them.”
Of course, his company stands to benefit from this nascent trend, but there are other reasons that this could be a positive for AI advancement.
Moving Beyond English
English is currently and unsurprisingly the dominant language in AI. “An estimated 90 per cent of training data for current generative AI systems stems from English,” writes Celeste Rodriguez Louro, linguistics professor at the University of Western Australia. As developers run out of quality human-produced English-language content on which to train new models, content in other languages might do more than just provide at least a temporary solution, and that’s before we consider the likelihood that some human knowledge has yet to be translated into English and thus, yet to be incorporated into existing AI models.
Let’s be clear: The language of William Shakespeare needs no defense whatsoever. Nevertheless, language is thought—a reflection of how people in different countries and cultures view and categorize the world. That’s why some words have no true equivalent in English and must be expressed in their original language: schadenfreude (German), hygge (Danish), and wabi-sabi (Japanese), for instance.
Furthermore, multilingual people often assert that learning a new language can enhance the brain’s ability to understand some ideas and process information. As one simplistic example, Russian speakers have been shown to distinguish between shades of blue more quickly than English speakers. Researchers believe this is linked to the fact that the Russian language has specific words to describe lighter shades of blue (goluboy) and darker shades (siniy).
Many researchers argue that if this is true for human brains, then it might be true for AIs as well. Some hypothesize that AI models trained in other languages could bring different advantages and capabilities to the table, and that’s before considering that AIs geared to speakers of different languages are likely to help expand the market for AIs in general.
In Europe, Nvidia and Google are particularly active in working to help countries develop sovereign AIs, with partnerships in Germany, France, Italy, Spain, and beyond. Nvidia is also active in Japanese and Vietnamese efforts along these lines.
Chinese tech companies, particularly Zhipu AI and Huawei, are also eager to play a role in helping others develop non-English sovereign AIs, with an unsurprising focus in Asian countries such as Malaysia and Singapore. Saudi Arabia is noteworthy in that the country has an array of partnerships with both U.S. and Chinese companies (Google, AMD, Amazon, Zhipu, and Huawei).
Looking Forward
Back in the early dot-com days, few accurately predicted which companies would ultimately come up with truly groundbreaking ways to use the capabilities of the Internet. The consensus view was that companies like AOL, Amazon, and Yahoo would be the new titans, with companies like Lucent Technologies, Cisco Systems, Qwest, and Level 3 Communications helping on the infrastructure side of things. Yet today, many of these names no longer exist as independent companies – if they even exist as entities at all. While Amazon has fulfilled its promise and is regarded as one of the Magnificent Seven, today’s Amazon — a tech, e-commerce, and entertainment powerhouse with expertise in logistics and supply chain management — is very different from its first incarnation as a mere online bookseller.
Similarly, some small, little-known company currently exploring a somewhat mundane use for AI might someday discover new routes on its way to magnificence. For that reason, today’s discussion provided a brief exploration of some not-as-glaringly-obvious companies that are experimenting with what can be done with AI. Ultimately, it might be time to look beyond who’s developing AI technology and devote some attention to what other businesses are doing with it.
That doesn’t make the investment case for today’s mainstays – chip companies like Nvidia and AMD, along with companies like Microsoft, Amazon, Alphabet, and Meta – less compelling, but it does suggest that we might now be entering a stage in which a few newcomers might find a way to ascend the mountain.
As always, Signal From Noise should not be used as a source of investment recommendations but rather ideas for further investigation. We encourage you to explore our full Signal From Noise library, which includes deep dives on the military drone industry, the presidential effect on markets, the America First trade, ChatGPT’s challenge to Google Search, and the rising wealth of women. You’ll also find discussions about consumer stocks, the TikTok demographic, and weight loss-related investments.
