2025 was the year that artificial intelligence was examined

Money was no object for the AI ​​industry in early 2025. Vital testing crept in in the second half of the year.

OpenAI raised $40 billion at a $300 billion valuation. Safe Superintelligence and Thinking Machine Labs raised individual $2 billion seed rounds before shipping a single product. Even first-time founders were raising money on a scale previously reserved only for big tech companies.

Such astronomical investments were followed by incredible spending. Meta spent nearly $15 billion to lock up Scale AI CEO Alexander Wang and has spent countless millions poaching talent from other AI labs. Meanwhile, the biggest AI players have promised nearly $1.3 trillion in future infrastructure spending.

The first half of 2025 matches investor enthusiasm and interest in the previous year. This mood has changed in recent months to provide a vital check of sorts. Extreme optimism about artificial intelligence, and the wild valuations that accompany it, remains. But this rosy outlook is now tempered by concerns about the bursting of the AI ​​bubble, user safety, and the sustainability of technological progress at the current pace.

The era of unabashed acceptance and celebration of AI is fading with just a stroke of the edges. And with it more scrutiny and questions. Can AI companies maintain their speed? Will scaling in the post-DeepSeek era require billions? Is there a business model that returns a fraction of the billions invested?

We have been there every step of the way. And our most popular stories of 2025 tell the real story: an industry that delivers reality testing even as it promises to reshape reality itself.

How the year started

The largest AI labs got even bigger this year.

In 2025 alone, OpenAI raised a $40 billion round led by SoftBank at a post-cash valuation of $300 billion. The company also reportedly has investors like Amazon circling computing-related circular deals and is in talks to raise $100 billion at a valuation of $830 billion. evaluation. This would bring OpenAI closer to the $1 trillion valuation it is said to be seeking in its IPO next year.

The OpenAI Anthropic competition also raised $16.5 billion this year in two rounds; Its recent increase has raised its valuation to $183 billion, with the participation of large companies such as Iconiq Capital, Fidelity, and the Qatar Investment Authority. (CEO Dario Amodei admitted to employees in a leaked memo that he was “not happy” about taking money from the dictatorial Gulf states.)

Then there’s Elon Musk’s xAI company, which raised at least $10 billion this year after acquiring Company X.

We’ve also seen smaller, newer startups get a big push from foaming-mouthed investors.

Former OpenAI CTO Thinking Machine Labs has secured a $2 billion seed round at a $12 billion valuation despite sharing almost no information about its product offerings. $200 million Vibe coding startup Lovable got the horn of a unicorn just eight months after launching; This month, Lovable raised another $330 million at a post-cash valuation of nearly $7 billion. We cannot ignore the artificial intelligence startup Mercor, which raised $450 million this year across two rounds, bringing its latest valuation to $10 billion.

These ridiculously huge valuations are still happening even against the backdrop of still-modest company adoption numbers and serious infrastructure constraints, adding to fears of an AI bubble.

My son, my dear, my son

For big companies, these numbers don’t come out of nowhere. Justifying these assessments requires building massive amounts of infrastructure.

The result has created a vicious circle. Capital raised to finance computing is increasingly tied to deals in which the same money flows back into chips, cloud contracts, and energy, as seen in OpenAI’s infrastructure-related financing with Nvidia. In practice, it blurs the line between investment and customer demand, raising concerns that the AI ​​boom is powered by the circular economy rather than sustainable use.

Some of the biggest deals this year that fueled the infrastructure boom were:

• Stargate, a joint venture between SoftBank, OpenAI, and Oracle, which involves up to $500 billion to build AI infrastructure in the United States.
• Alphabet’s acquisition of power infrastructure and data center provider Intersect for $4.75 billion, which comes as the company said in October that it plans to raise its computing spending in 2026 to $93 billion.
• Meta’s accelerating data center expansion has pushed its projected capital expenditures to $72 billion in 2025 as the company races to secure enough compute to train and run next-generation models.

But cracks are starting to appear. Private financing partner, Blue Owl Capital, recently pulled out of a planned $10 billion Oracle data center deal tied to OpenAI capacity, underscoring how fragile some of these capital stacks are.

Whether all this spending will eventually materialize is another question. Grid constraints, rising construction and energy costs, and growing resistance from residents and policymakers — including calls from figures like Sen. Bernie Sanders to rein in data center expansion — are already slowing projects in some areas.

Even as investment in AI remains massive, the reality of infrastructure is starting to dampen the hype.

Re-prediction

In 2023 and 2024, every major model release was a revelation, with new capabilities and new reasons to get caught up in the hype. This year, the magic wore off, and nothing embodied this shift better than the launch of OpenAI’s GPT-5.

Although it made sense on paper, it didn’t land with the same impact it did Previous versions such as GPT-4 and 4o. Similar patterns have emerged across the industry where improvements made by LLM providers have been less transformational and more incremental or domain-specific.

Even Gemini 3, which tops many benchmarks, was only a major accomplishment insofar as it put Google back on par with OpenAI — which sparked Sam Altman’s infamous “Code Red” memo and OpenAI’s battle to maintain dominance.

There has also been a reset this year in terms of where we expect frontier models to come from. DeepSeek’s launch of R1, its “logical” model that competed with OpenAI in key benchmarks, demonstrated that new labs can ship credible models quickly and at a fraction of the cost.

From model breakthroughs to business models

As each jump between new models shrinks in size, investors focus less on the model’s raw capacity and more on its surroundings. The question is: Who can turn AI into a product that people depend on, pay for, and integrate into their daily workflow?

This shift is manifesting itself in many ways as companies see what works, and what customers will let fly. For example, Perplexity, an AI research startup, briefly toyed with the idea of ​​tracking users’ online movements to sell them highly personalized ads. Meanwhile, OpenAI was reportedly considering charging up to $20,000 a month for specialized AI, a sign of how aggressively companies are testing what customers might be willing to pay.

But more than anything else, the battle has moved to distribution. Perplexity is trying to stay relevant by launching its own Comet browser with proxy capabilities and paying Snap $400 million to power search within Snapchat, effectively buying its way into existing user paths.

OpenAI is pursuing a parallel strategy, expanding ChatGPT beyond chatbots and into the platform. OpenAI has launched its own Atlas Browser and other consumer-oriented features like Pulse, while also courting enterprises and developers by launching apps within ChatGPT itself.

Google, for its part, relies on filling jobs. On the consumer side, Gemini is integrated directly into products like Google Calendar, while on the enterprise side, the company hosts MCP connectors to make its ecosystem more difficult to displace.

In a market where it has become difficult to differentiate by abandoning a new model, owning the customer and business model is the real moat.

Verify trust and safety

AI companies have come under unprecedented scrutiny in 2025. More than 50 copyright lawsuits have been filed in the courts, while reports of “AI psychosis” – the result of chatbots promoting delusions and allegedly contributing to multiple suicides and other life-threatening events – have sparked calls for trust and safety reform.

While some copyright battles have ended — such as Anthropic’s $1.5 billion settlement for authors — most remain unresolved. However, the conversation appears to be shifting from resisting the use of copyrighted content for training to demanding compensation. (See: New York Times sues Perplexity for copyright infringement.)

Meanwhile, mental health concerns around AI chatbot interactions — and their ingratiating responses — have emerged as a serious public health issue following multiple deaths by suicide and life-threatening delusions in teens and adults after prolonged use of chatbots. The result has been lawsuits, widespread concern among mental health professionals, and swift policy responses like SB 243 in California regulating AI companion robots.

Perhaps most tellingly, the calls for restrictions are not coming from the usual anti-technology suspects.

Industry leaders have warned against the “engaging interactivity” of chatbots, and even Sam Altman has warned against too much emotional reliance on ChatGPT.

Even the laboratories themselves began raising alarms. Anthropic’s May safety report documented Claude Opus 4’s attempt to blackmail engineers into preventing its shutdown. Subtext? Scaling without understanding what you’ve built is no longer a viable strategy.

Looking forward

If 2025 is the year AI starts to grow and face difficult questions, then 2026 will be the year it has to answer them. The hype cycle is starting to wear off, and now AI companies will have to prove their business models and demonstrate their true economic value.

The era of “trust us, the payoff will come” is coming to an end. What comes next will either be a vindication or a reckoning that makes the dot-com collapse look like a bad day for Nvidia trading. It’s time to place your bets.