The financial markets, including the companies that bet everything on a certain two-letter acronym called artificial intelligence (AI), noticed a shift in 2025. A few years ago, investors were in a tizzy over AI’s role in the 21st-century global economic order, and companies could receive a valuation boost simply by mentioning AI in their earnings calls. However, this year, it is taking more than hype and big growth numbers to grab investor attention, with earnings growth becoming a crucial factor.
The initial burst of excitement around AI, which lasted from 2023 to 2024, was fueled by hype and billions of dollars in inflows (hundreds of billions of dollars) that many companies rode on without the profits to back it up.
The seven companies that did not need blockbuster profits to draw matching inflows, rocketing valuations based on future potential rather than current performance, are collectively referred to as the Magnificent Seven, which include Microsoft, Google parent Alphabet, Tesla, Amazon, Apple, Facebook parent Meta, and NVIDIA.
So, what connects the first six companies? Other than Apple (which spends a minuscule amount on NVIDIA hardware), they are all major customers of NVIDIA, buying boatloads of the semiconductor giant’s chips to train their large language models (LLMs) and power their AI. This is why NVIDIA ignited the AI rally, and the rally came with gains.
Shares of NVIDIA rose 239% in 2023 and another 171% in 2024, yet another strong year, but the heavyweight tech name has had difficulty finding traction so far in 2025. When the company posted 114% annual revenue growth at the beginning of the year, it was not enough to get traders and investors running to load up on the shares. In 2024 (fiscal 2025), the company reported $130.5 billion in revenue, which was more than double the previous year’s figure of $60.9 billion.
Investor response was a smattering of polite claps, yawns, and just enough buying momentum to keep NVIDIA’s shares relatively flat in the days after the February 26 earnings report that included fiscal 2025 performance. The muted market response might indicate that even the highest expectations had already been built into the price, or that investors are becoming less interested in fundamentals and more enamoured with the notion of stratospheric growth, given that NVIDIA blew out expectations by 265% year-over-year in Q4 2023.
When it comes to pumping their money into AI, investors are now looking at fundamentals like sustainable margins, monetisation strategies, and disciplined capital spending. Companies that can translate AI innovation into reliable, long-term profits will be the winners. Meanwhile, others may find it difficult to justify their higher valuations in a market focused on earnings. Some spending must happen before those earnings can take effect and, of course, be realised.
Capex: Hero or villain?
Capital expenditures (capex) are the money a company allocates to investing in innovation, upgrades, and new assets, such as hardware or software. For example, in the AI world, companies tend to spend on hardware and data centres to support high-performance computing.
The other thing is that capex tends to be much more unpopular with investors because the latter, again, want to see value today, and they want to see value in the next few months. They do not want to see value in the next year, let alone the next decade. With $80 billion in capital expenditure, 2025 has been a year of bold investments, with the “Magnificent Seven” showing their commitment to the future of AI.
Microsoft has invested $80 billion to grow its data centres and AI infrastructure to power its Azure cloud platform and its broader enterprise ecosystem, and its AI chatbot, Copilot, will become a standard part of the toolkit for businesses and consumers to streamline workflow and day-to-day tasks.
Microsoft is at the forefront of generative AI after it backed OpenAI, the parent of ChatGPT, with a 49% stake, thanks to a $13 billion bet, and Alphabet, Google’s parent, has pledged around $75 billion to similar efforts, bolstering its role in AI research and cloud services, much of which is going towards Gemini, Google’s generative AI model.
Amazon is making the biggest bet of all, with capital expenditures over $100 billion for 2025, and much of that will go into AI infrastructure for Amazon Web Services (AWS), the core of its enterprise operations and a key profit driver.
Meta has also sharply increased its guidance to a total of $60 billion to $65 billion, a nearly 70% increase from earlier estimates. Most of that big-ticket spending will be for warehouse-sized data centres to run the AI products across its apps, such as Facebook, Instagram, and WhatsApp.
The other “Magnificent Seven” members, Apple, Tesla, and NVIDIA, have not announced their capex plans, but their forward guidance and spending levels indicate ongoing, large investment in AI and related technologies, such as “Apple Intelligence,” which is the tech giant’s catch-up AI effort; Full Self-Driving (FS) by Tesla, its highest-level driver assistance software; and “Blackwell,” the next big thing in GPU and chip manufacturing for NVIDIA.
When viewed collectively, this investment wave in 2025, representing over $300 billion among the top players alone, is more than an optimistic note and represents a fundamental change in how value will be created in the next decade.
These firms are not backing down on their bold investment initiatives, despite market volatility, occasional pushbacks from investors, and macroeconomic challenges. They are investing today for tomorrow’s growth, knowing that the returns might not be immediate.
In AI 1.0, markets paid for guidance and expectations, but in AI 2.0, they pay for performance. The speculative phase is over, replaced by operational discipline and value creation based on the implementation of new technology.
With high interest rates compared to four years ago, the notion that capital has a cost has been reintroduced, and the focus has returned to those who have the upper hand, namely the big infrastructure players with pricing power and established supply chains.
What to expect in 2026
After a couple of years of heady share-price gains followed by a frenetic race to build out infrastructure, the market is now moving on to a new phase, one of execution, efficiency, and results.
Looking forward to 2026, three trends will be key to the AI earnings cycle, and enterprise adoption will be the true test. Is it being paid for at scale? Are workflows changing in ways that are both significant and monetizable? Do consumers need AI daily? These questions will require quick answers.
Energy prices rise, and infrastructure costs are high. The leaner, more efficient companies will be able to hold the line on profitability.
Competitive moats will matter, and by 2026, investors will need to know: Who owns the data? Who controls distribution? Who has proprietary models, scale advantages, or ecosystem lock-in? As the space matures, staying power, not just innovation, will differentiate the leaders from the waning hype.
Moreover, the companies that demonstrate resilience will be those capable of converting their massive AI investments into tangible, revenue-generating products and services. While the early phase of the AI boom rewarded ambition, the next phase will reward operational precision.
Investors will scrutinise not only how much companies spend, but how efficiently those dollars translate into ecosystem advantages, customer retention, and recurring revenue models. The winners will distinguish themselves through strategic discipline by balancing cutting-edge research with commercial clarity, securing critical infrastructure partnerships, and maintaining supply chains that can withstand global uncertainty.
Also, governments are moving toward stricter oversight of AI training data, model transparency, and the environmental impact of data centre expansion. Companies that anticipate these shifts, incorporating compliance into their core strategies rather than treating it as a last-minute obligation, will navigate the landscape with fewer disruptions and lower long-term costs.
Traditional tech giants will no longer be the only ones capable of delivering advanced AI solutions. Leaner, specialised firms may carve out niches in sectors such as healthcare, manufacturing, and cybersecurity. In this environment, adaptability becomes as critical as scale, and companies unable to evolve quickly will risk losing relevance despite earlier advantages.