After a year of rapid, AI-driven growth in the technology sector in 2025, momentum carried into early 2026, with the S&P 500 repeatedly approaching the 7,000 level. At the same time, a wave of earnings releases from major technology companies—including Intel, Meta, Microsoft, Apple, AMD, Alphabet, Qualcomm, Amazon, and Tesla—came into focus.
As the new year began, the most widely discussed topic in the market was the efficiency of capital deployment by big tech companies. Based on recent earnings calls, expectations for massive capital expenditures (CapEx) have not translated into higher valuations or stronger market optimism. Instead, post-earnings share price declines have occurred repeatedly. That said, the underlying financial performance of these companies has undeniably been strong, with AI-driven revenue growth clearly visible.
Beyond AI infrastructure providers, new breakthroughs in AI end-user applications have emerged in the new year. In particular, Anthropic—led by former OpenAI engineers Dario and Daniela Amodei—has entered the AI applications market with its Claude model series, prompting a reassessment of valuations across SaaS software stocks. Early 2026 has already been eventful. This article explores what has been happening in equity markets at the start of the year, how AI may further influence U.S. stock performance, and how a new wave of AI challengers could disrupt incumbent winners.
AI Giants’ Capital Expenditure Expansion Raises Bubble Concerns
In reality, capital spending by AI leaders is not a new topic. Since the early stages of AI development, the market has harbored concerns about capital allocation risks. Recently, however, AI-related capital expenditures have reached unprecedented levels. Alphabet, Microsoft, Meta, and Amazon alone—the four major cloud providers—are expected to post combined capital expenditures approaching US$700 billion in 2026, representing year-over-year growth of more than 60% and accounting for roughly 2.1% of U.S. GDP. This scale surpasses that of the Apollo moon landing program and the construction of the U.S. interstate highway system.
| Unit: US$100 million | Alphabet (Google) | Meta | Microsoft | Amazon | Apple | Tesla |
|---|---|---|---|---|---|---|
| CY25 Capital Expenditure | 914 | 722 | 1,180 | 1,310 | 121 | 85 |
| CY26 CapEx Guidance | 1,750–1,850 | 1,150–1,350 | 1,100–1,200 | 2,000 | 130–135 | 200 |
| CapEx YoY | 102% | 87% | 2% | 53% | 12% | 135% |
| CY25 Operating Cash Flow (OCF) | 1,647 | 1,158 | 1,605 | 1,395 | 1,355 | 147 |
| CY25 CapEx as % of OCF | 55% | 62% | 74% | 94% | 9% | 58% |
| CY25 CapEx as % of Revenue | 23% | 36% | 39% | 18% | 3% | 9% |
| CY25 Free Cash Flow (UFCF) | 385 | 242 | 553 | 252 | 1,063 | 39 |
| Key Investment Focus | Gemini 3 Pro, TPU v9, data center expansion | Llama 4, Reality Labs (metaverse) | Azure AI expansion, OpenAI Stargate partnership | Trainium 2/3, Amazon Leo (Project Kuiper), Project Rainier | Apple Intelligence | Investment in six new plants (refining, Optimus, Cybercab, Semi, AI computing, megafactory, etc.) |
Apple and Diverging CapEx Strategies Among Tech Giants
Amid massive capital outlays, investors are increasingly focused on capital efficiency. Among the six companies above, although all have raised their 2026 CapEx guidance, their AI-era strategies differ significantly. Apple stands out, with capital expenditure accounting for just 9% of operating cash flow in 2025—well below peers. This suggests that Apple continues to avoid aggressive investment in data centers and in-house chip development. At the end of January, Apple announced a partnership with Google to integrate Gemini into Siri on the iPhone.
For 2026, the market expects Apple’s CapEx to remain low at around US$13 billion, largely to maintain existing operations. By contrast, Google, Meta, Microsoft, and Tesla reported 2025 CapEx levels equivalent to roughly 55%–75% of operating cash flow, eroding more than half of free cash flow. Google, Meta, Tesla, and Amazon have also significantly increased their 2026 CapEx plans. If earnings growth fails to keep pace in the coming quarters, these companies could face negative free cash flow. Notably, Tesla’s latest guidance does not yet include planned investments in solar energy and semiconductor facilities, leaving room for further upward revisions.
Amazon Faces Severe Cash Flow Pressure
Among the tech giants, Amazon’s capital spending is the most aggressive, with 2025 CapEx nearly equal to operating cash flow. As a result, free cash flow has shrunk by roughly half compared with 2024, leading the market to expect a high likelihood of negative free cash flow from 2026 onward.
Unlike Microsoft, Meta, or Google, which rely heavily on high-margin software and advertising, more than half of Amazon’s revenue comes from low-margin e-commerce retail. Investing heavily in AI infrastructure on a comparatively weaker profit base exposes Amazon to greater cash flow and financial risk than its peers.
In terms of allocation, while most companies focus CapEx on servers and AI computing infrastructure, Amazon’s US$200 billion plan also includes the Amazon Leo initiative. Formerly known as Project Kuiper, Amazon Leo is a low-earth-orbit satellite project that will place significant pressure on near-term profitability and free cash flow. During its earnings call, Amazon noted that the satellite program alone is expected to add roughly US$1 billion in year-over-year operating income costs in the first quarter.
Accelerating AI Revenue Growth Fuels Rising Market Expectations
This earnings season has clearly highlighted AI-driven profit momentum, with Meta posting the highest year-over-year revenue growth. AI has significantly improved advertising efficiency, driving strong growth in its core advertising business and pushing revenue and profits well above market expectations. Meta’s shares surged more than 11% in after-hours trading. Other tech giants also delivered solid results, with many exceeding consensus forecasts. Still, aside from Meta, most companies saw share price pullbacks after earnings releases.
| Unit: US$100 million | Alphabet (Google) | Meta | Microsoft | Amazon | Apple | Tesla |
|---|---|---|---|---|---|---|
| CY25 Revenue | 4,028 | 2,010 | 3,055 | 7,169 | 4,356 | 948 |
| CY24 Revenue | 3,500 | 1,645 | 2,618 | 6,380 | 3,958 | 977 |
| Revenue YoY | 15% | 22% | 17% | 12% | 10% | (3%) |
A clear trend has emerged in early 2026: even when AI leaders significantly outperform both company guidance and market consensus, share prices do not necessarily follow. This suggests that valuation standards for semiconductor and AI-related companies have risen sharply.
As AI technology advances rapidly and capital investment continues to grow, the semiconductor industry—central to AI infrastructure—has been assigned exceptionally high long-term growth expectations. Investors increasingly view GPUs, CPUs, advanced manufacturing, memory, and data centers as direct beneficiaries of the AI wave, leading to ever-higher expectations for revenue growth, margin expansion, and medium- to long-term outlooks. As a result, results that merely meet or slightly exceed expectations may be interpreted as insufficient, implying that future AI growth has already been priced into current valuations.
Apple is also worth noting. In 2025, Apple delivered strong revenue growth, driven largely by a 38% year-over-year surge in Greater China revenue in the second half. iPhone sales reached a record single-quarter high in China. Despite early-year challenges from rising memory prices and tight supply, Apple’s high-margin structure and large-scale procurement limited the impact relative to Android competitors. This underscores how Apple’s premium positioning has further strengthened its pricing power. Importantly, while achieving revenue growth comparable to peers, Apple avoided heavy capital spending and instead enhanced product value through partnerships, giving its stock more defensive characteristics amid rising CapEx concerns.
AI Agents and the Repricing of SaaS Valuations
From late 2025 into early 2026, several notable developments emerged at the AI application layer, including discussions around Physical AI, the rise of Anthropic’s Claude models, and the rapid popularity of OpenClaw.
- Physical AI is moving beyond pure text prediction and abstract applications, beginning to learn physical laws such as gravity, fluid dynamics, and material mechanics. This enables more accurate simulation of real-world systems and is expected to have far-reaching implications for robotics automation and physical system control.
- Anthropic’s Claude models have demonstrated remarkable logical rigor in software development, capable of independently breaking down and executing complex, multi-step engineering tasks, making them indispensable tools for developers.
- OpenClaw (formerly Clawdbot), which has quickly gained traction in the open-source community, further expands AI agent applications. It can remotely control a user’s computer via platforms like Telegram and WhatsApp, handling tasks such as email organization, coding, and even facilitating collaboration among multiple AI agents. Its ability to run locally with long-term memory pushes end-user applications toward true personalization and autonomy.
Anthropic Challenges the SaaS Industry’s Moat
Led by former OpenAI engineers Dario and Daniela Amodei, Anthropic’s direct push into the AI application layer with its Claude models has become a key factor behind the recent sharp correction in SaaS software stocks. In early 2026, global software equities experienced severe volatility. The software sector index recorded its worst single-month performance since 2008 in January, falling about 15%. Individual stocks such as Salesforce and ServiceNow were down more than 20% year to date, while Adobe, SAP, and Snowflake at one point approached 20% declines. Figma, which went public in 2025, plunged as much as 41%.
AI Agent Automation Lowers the Barrier to Software Development
The core reason behind the sell-off is concern over whether AI—particularly Anthropic’s next-generation assistant Claude Cowork and automated legal and data tools—could replace existing software products.
AI agents can directly read and write documents and analyze data, potentially reducing the need for companies to purchase expensive SaaS licenses for every employee. As AI coding capabilities advance rapidly, enterprises may prefer to build customized internal tools rather than paying millions of dollars annually to third-party software providers. The rise of AI end-user applications directly challenges the moats of traditional SaaS companies. If engineers no longer need Jira and creative professionals no longer need to master Photoshop, the future of these software companies could be at risk. As such, investors must reassess whether SaaS companies’ competitive advantages still exist—and whether they could ultimately be displaced by AI.
Brief Commentary
Since the launch of OpenAI’s ChatGPT, AI has evolved from a focus on computing power accumulation to commercial validation. Along this journey, markets have shifted from exuberant capital spending toward more cautious discussions around return on invested capital. As a result, discussions around AI in 2026 are likely to continue centering on cash flow sustainability and the cost of capital.
Earnings results clearly show that AI is contributing meaningfully to revenue, but they also reveal big tech’s persistent fear of missing out on AI computing power, driving continued heavy investment. For investors, identifying companies that can effectively balance capital expenditure with investment returns will be the key challenge in the next phase.
