The third quarter’s earnings reports have already generated the narrative we’ll be talking about well into 2027: Can big tech’s core business survive through the rapidly changing landscape of AI?
At the time of this writing, four tech giants – Meta, Microsoft, Alphabet (Google), and Amazon have released their Q3 earnings. The key takeaways? Spending remains extremely elevated, and core businesses are showing signs of pressure.
Each company is handling this pressure differently, and some are better positioned to weather shifts in the landscape more than others. It’s important we understand these variables given the fact that just these four companies alone represent almost 17% of the S&P 500 index. Let’s dive into some of the details we’ve seen emerge throughout the Q3 earnings season thus far.
The Capital Spending Machine: Why Equity Is Taking the Lead
Big tech capital expenditures—money spent on physical infrastructure like high-end graphics chips, fiber cabling, and server facilities—surpassed unprecedented milestones this quarter. Microsoft, Alphabet, Amazon, and Meta have signaled that their infrastructure investments will remain elevated well into next year.
What stands out is how companies are funding this buildout. While tech behemoths traditionally relied on cash reserves and occasional corporate bond issuance, the financial landscape has shifted. Interest rates remain a factor, making massive long-term corporate debt expensive. More importantly, investing tens of billions of dollars into fast-depreciating computer chips is inherently risky. If a company takes on fixed bond payments to buy chips that become obsolete in two or three years, a sudden slowdown in demand could crush its balance sheet.
As a result, financing across the AI ecosystem—especially among frontier labs, cloud providers, and infrastructure joint ventures—is tilting heavily toward equity raises. Selling ownership stakes or issuing stock to strategic partners spreads the technological risk. Lenders demand reliable, cash-generating collateral, whereas equity investors are willing to absorb volatility in exchange for upside potential. By leaning into equity rather than piling on debt obligations, companies avoid burdening their core operations with mandatory interest payments while building out unproven capacity.
The critical theme within this strategy is how long will the equity market continue to willingly fund big tech’s cash needs? Meta recently conducted a stock issuance of $25 billion which was five times oversubscribed. Which means for every share they issued, 5 investment firms were waiting to purchase it. This is a positive sign for investors who believe the AI trend can continue at its current trajectory. But as we all know, these trends can shift quickly, and if the bond market is already too expensive for new issuance, a lack of appetite for new equity issuance could put the brakes on further AI development.
Pressure on the Cash Cows: Google and Intel
While cloud divisions are expanding, legacy cash-generating engines are feeling the strain of the AI transition.
Google’s Core Search Business
For two decades, Google’s primary search engine operated with extraordinarily high profit margins. A simple text search cost fractions of a cent to execute, and sponsored links generated steady revenue. Generative search summaries and conversational queries run on complex neural networks that consume substantially more computing power per prompt than traditional index searches. At the same time, conversational AI platforms are changing browsing habits, threatening traditional search ad clicks.
Alphabet is effectively spending more computing power per query while defending its digital advertising crown. The key will be to find a way to translate their core advertising margins to generative search.
Intel and the PC Market
Intel faces a different challenge: component inflation in personal computers. Hardware makers are rushing to brand their laptops and desktops as “AI PCs,” integrating dedicated neural processing units (NPUs) alongside faster, high-bandwidth memory. These specialized components have driven manufacturing costs upward. Consumers and enterprise buyers, facing higher sticker prices for hardware that does not yet show distinct day-to-day advantages over standard PCs, are holding off on upgrades. As hardware costs rise, Intel’s mainstream PC client division faces pinched margins and cautious consumer demand.
The Token Paradox: Soaring Volume, Crashing Prices
The fundamental metric of generative AI is the token—the basic chunk of characters or words a computer model processes. Every time you use AI, you’re “spending” tokens to create an answer or image. At the enterprise level, companies will purchase a certain amount of tokens to use each month. For retail consumers, we will pay a monthly fee to access AI tools (or just use the free versions), and this fee (or advertising within the free versions) pays for the tokens used to generate your results.
Across every benchmark, total token consumption reached all-time highs this quarter. Enterprises are automating customer support, writing software code, and indexing internal documents at scale.
Despite surging demand, the cost per token has plummeted drastically. Data based on industry benchmarks across multiple mainstream providers shows token costs going from $30/token in 2023 down to $0.20/token in 2026. This swift disinflation is great for consumers, but places significant strain on developers as their margins come under pressure as a result of falling revenue per token.
Intense competition among frontier developers—including OpenAI, Anthropic, Google, and open-source models like Meta’s Llama family—has sparked an aggressive price war. Furthermore, chip efficiency and architectural breakthroughs, such as specialized small-parameter models, have lowered the compute required to generate each word.
This creates a split dynamic: for developers and startups building applications, intelligence is becoming remarkably affordable. But for model creators, generating revenue requires massive, compounding transaction volumes just to offset falling per-unit prices.
Looking Ahead to the Close of 2026
As markets head toward the final quarter of the year, investors and industry watchers will be focusing on three primary indicators:
- Capital Expenditures Real Return on Capital: Scrutiny will intensify on software vendors to prove that enterprise pilot programs are converting into lasting, recurring software licenses, rather than one-time experiments.
- PC and Device Refresh Cycles: Holiday retail performance will indicate whether everyday consumers will pay a premium for AI-enabled hardware, or if manufacturers will need to discount inventory.
- Token Costs: As model token prices continue to compress, the market will monitor whether cloud providers can sustain operational profitability through sheer usage volume.
These trends, in addition to higher interest rates, inflationary pressures, and Anthropic’s IPO will all be closely monitored to gauge how the tech industry and consumers are responding to these rapidly changing dynamics.
