AI, cloud, and communications infrastructure was once again a focal point of the investment markets in 2025. Investors feasted on the story of unlimited demand for cloud infrastructure to fuel AI and related apps—to the tune of trillions of dollars.
While in the near term our data shows that it’s likely that the pace of spending on cloud infrastructure will continue to be brisk, the story gets more complicated in 2026. Investors will likely grow more discerning as they look out to spending 2027 and start asking questions about profits. Some of the largest hyperscaler clouds are spending a larger share of their revenues and profits on new data centers that have yet to prove a return on investment. Questions are also being raised about the amount of debt being raised to build out this infrastructure.
Looking back, the stocks that get all the headlines weren’t necessarily the biggest winners. For example, Cloudflare is the biggest winner we tracked—a bit off the radar of the mainstream. Let’s look at some of the best annual results in stock prices of the notable public companies we follow (data as of December 30):
Cloudflare (NET): +80%
Alphabet (GOOGL): +65%
Nvidia (NVDA): +40%
Cisco (CSCO): +30%
Arista Networks (ANET): +21%
Oracle (ORCL): +19%
Microsoft (MSFT): +16%
HPE (HPE): +10%
Dell (DELL): +7%
Amazon (AMZN): +2%
Palo Alto Networks (PANW): +2%
Now, let’s dive into some of the most interesting stories.
The Top AI Infra Investment Stories
One sign that winning won’t always be easy can be seen in what happened to Oracle, which expressed a big shift in AI sentiment in the latter half of 2025. Let’s take a look at that and at the other top AI infrastructure stories in 2025.
Oracle Rises, then Falls. Oracle’s spectacular rise and fall ranks up there with one of the strangest stories of the year. Things reached fever pitch in September. As part of the Stargate project, Oracle pledged 4.5 GW of AI compute capacity to OpenAI in a deal worth $300 billion—though this “deal” was nothing more than a memorandum of understanding, with no firm long-term commitment. Shares vaulted as high as $350 in September, but the stock has since sunk back to $196—a loss of 40% since that announcement. The big problem is that these deals are now shifting to being financed with debt. Lately, there are reports that one of Oracle’s large backers, Blue Owl Capital, is getting cold feet on concerns about Oracle’s mounting debt and AI spending. In addition, investors recently questioned the information from Oracle’s SEC filings that it has $248 billion in future data-center lease payments on the books, which is a huge liability.
More than anything, the Oracle rollercoaster is indicative of the AI sentiment as a whole. We don’t know how profitable these data centers will be, and neither do investors.
The Market Gets Dangerously Concentrated on OpenAI. The AI infrastructure spend is largely speculative, with investors betting that trillions of dollars on new AI data centers will result in enterprise and consumer demand, driving revenues and profits for AI companies. There’s only one problem: This return on investment isn’t evident yet. A single private company, OpenAI, is central to the largest AI capacity orders being placed with Nvidia and Oracle, but the company is losing money and it either needs to increase its growth rate (unlikely), or raise more money to the tune of close to a trillion dollars. Toward the end of the year, some of the air came out of the AI trades as investors grew concerned about the role of OpenAI in the AI market, given its position in a circular pattern of spending in which Nvidia, Oracle, CoreWeave, and others are funneling billions into data-center buildouts on pledges from OpenAI. Yet OpenAI is reportedly on track to burn an estimated $8.5 billion in cash this year despite expecting to score $13 billion in annual sales. And it doesn’t anticipate profitability until 2029 or 2030. More importantly, OpenAI is a private company, so there is a lack of transparency into its business.
The U.S. Government Bets on Tech. In one of the stranger twists of politics, the Trump administration, which favors deregulation, has become a bigger influence on the markets. U.S. President Donald Trump has prioritized bringing more technology products, including chip production, back to America. The Stargate project, announced in January 2025, put OpenAI, Oracle, and SoftBank in the spotlight as leading developers of a $500 billion planned data-center buildout. In August, the government pledged to invest billions in Intel. It’s also making big bets on critical metals, such as lithium and cobalt, which are critical to technology. At the same time, it’s favored a deregulatory approach to AI and data-center buildouts, which is controversial and potentially dangerous. In 2026, expect to see more government involvement in the tech sector, particularly in AI, where Trump recently announced the lifting of state and local restrictions on the technology. Look for more involvement from the adminstration, especially if things start to go sideways.
The Nvidia Mafia Gets Bigger. Call it a keiretsu or call it a mafia, but the business syndicates model has a long history in tech—going back to the “PayPal Mafia” and “WinTel.” This year, Nvidia expanded its influence with partners and customers by investing billions of dollars in a series of companies—not by acquiring them but by taking significant stakes. The roster includes Synopsys, in which Nvidia invested $2 billion; Intel, $5 billion; OpenAI, $100 billion; and Nokia, $1 billion. Nvidia also invested in numerous other companies, including Mistral AI, Wayve, and Reflection AI. In the most recent deal, Nvidia struck a $20 billion licensing deal with inference-chip company Groq. With friends like these, Nvidia is lining up technology (and friendly customers) to expand its influence in the AI ecosystem. Will it work? It’s probably smart business, but history has shown that technology markets often move faster and in more unpredictable ways than the dominant suppliers anticipate. Nvidia will have to stay on top of swift changes in the markets to make sure it doesn’t fall into the crossing-the-chasm trap.
Cisco Gets Its AI Act Together. In 2024, Futuriom pointed out that Cisco’s growth was static, as it was battling many challenges and losing its dominance in networking, especially in the key market of AI. It was losing market share to fast-moving technical companies such as Arista and Nvidia, and it was suffering from misdirection after dozens of acquisitions in many different markets. We noted that Cisco needed to focus on integrating recent acquisitions, rationalizing operating systems, and addressing declining core networking revenue, which saw a 23% drop in early 2025.
Cisco was either listening, or it learned that these problems needed to be addressed. Following a massive corporate restructuring at the end of 2024, Cisco put Jeetu Patel at the helm of Cisco’s Networking, Security, and Collaboration units. It pivoted toward better integration of its Nexus and Catalyst systems and pivoted toward AI infrastructure with a partnership with Nvidia. These moves have paid dividends, as Cisco’s stock had a strong year, rising 30% and hitting an all-time high. Recent earnings show the company has made enormous progress in addressing flaws in its strategy that we pointed out in June of 2024. In short, Cisco seems to have regained its footing in key markets, and with a strong balance sheet and conservative management culture, it’s a key AI technology player to watch in 2026.
Neoclouds and Alt-scalers Power Up. 2025 gave rise to the alternative cloud provider that specializes in renting accelerated computing infrastructure and GPUs for AI development. The subject of a recent report from Futuriom, the neoclouds have grabbed market share and gained investor attention. The surge spawned neocloud-related IPOs by CoreWeave and WhiteFiber, though market results have proved volatile. Futuriom is also watching alternative clouds (or “alt-scalers,” as we have recently taken to calling them), such as Cloudflare and Vultr, as they leverage their global infrastructure for enterprises looking for a more integrated approach to AI applications and security. As noted earlier, Cloudflare was a big winner on the investment front, up 85% on the year.
Hyperscalers Shift to Hybrid Models. AWS, Microsoft Azure, Google Cloud, and Oracle are acknowledging that most large customers are using more than one cloud supplier. While this trend has been building for a few years, AI will only accelerate the trend. As Deloitte research recently found, organizations are facing the critical challenge of scaling their computing infrastructure to keep up with growing use while also controlling rising costs. Multicloud connectivity needs are also growing, as evident when AWS and Google announced a mutual connectivity service at AWS re:Invent.
Futuriom’s own research highlights the growing benefits of hybrid cloud architectures, such as operational efficiency (see below).
Ethernet Wins in AI Networking. Futuriom research indicates that Ethernet is becoming the dominant, preferred network fabric for AI. While InfiniBand is still popular for specific back-end applications, momentum has grown in the Ultra Ethernet Consortium (UEC), with the largest players, including Arista Networks, Cisco, and Nvidia, adopting RoCE (RDMA over Converged Ethernet) to connect the most massive AI clusters as well as inference infrastructure. All the top networking players, including Arista, Cisco, and Nvidia, benefitted from the widespread adoption of Ethernet networks for diverse AI applications in 2025. We don’t see that ending in 2026.
Hyperscalers Grow Their Own Chips. Unwilling to rely on the whims of Nvidia and its suppliers, the leading cloud providers have developed (or are developing) their own chips, with some significant results. AWS’s Trainium has launched successfully. Anthropic has agreed to use up to 1 million Google TPUs. Chip competition is also growing with AMD. And a growing roster of startups are gathering like clouds on Nvidia’s horizon. Nvidia’s recent $20 billion licensing deal with Groq shows how that vendor is taking notice as the market is diversifying.
A More Volatile 2026?
If there’s one thing certain about markets, it’s that they can be unpredictable. 2025 started with a technology tariff panic and ended with AI euphoria. Yet, much of the enthusiasm has been based on uncertain terms, such as the trajectory of AI service adoption.
Futuriom’s research shows that AI adoption is broad based, but it’s still in the very early stages and is rife with challenges and questions, such as those about security and safety. It’s possible that the largest gains in AI are being seen by industrial and financial giants, as we have highlighted in our research. Companies such as John Deere, Eli Lilly, and Walmart also had very successful years, some of it aided by AI.
Investors will have to be nimble and stay on their toes in 2026, as the sentiment can shift quickly. We’ll be focused on how AI strategies can help optimize improvements across industries, which could be the best investment theme to watch in 2026. More on that in the coming weeks.
