Major technology companies are escalating their investments in artificial intelligence infrastructure at a staggering rate, but a closer look at their spending plans reveals starkly different strategies. While some giants like Microsoft are building capacity to meet immediate, contract-backed demand, others, including Meta, are making massive speculative bets on future, yet-to-be-defined AI services.
This divergence in capital expenditure highlights a critical split in the AI race: one path is paved with measurable, near-term returns, while the other ventures into a more uncertain, long-term vision. As spending forecasts for 2025 and 2026 continue to climb, investors are beginning to scrutinize which of these high-stakes gambles are most likely to pay off.
Key Takeaways
- Microsoft and Meta have announced that their AI capital spending will increase more in 2026 than in 2025, signaling a long-term investment cycle.
- Microsoft's spending is largely tied to existing customer demand and contracts, reducing immediate financial risk.
- Meta's investment strategy is more speculative, focused on developing future AI services for its massive user base, which has caused some investor uncertainty.
- Profitability is becoming a key metric, with depreciation costs from new data centers rising sharply across the industry.
- The nature of contracted revenue, known as RPO, varies significantly, with some deals representing concrete near-term income and others being more speculative long-term commitments.
The Two Roads of AI Investment
The latest earnings season has confirmed that the AI spending spree is not slowing down. Alphabet has increased its capital expenditure plans for the remainder of this year, while both Microsoft and Meta project that their spending growth will accelerate further into 2026. This sustained, high-level investment underscores the industry's belief that AI is the next major technological frontier.
However, the motivations behind these billions of dollars in spending are not uniform. Two distinct approaches are emerging: demand-driven expansion and speculative construction. This split is creating different risk profiles for the companies at the forefront of the AI revolution.
Microsoft's Demand-Driven Approach
Microsoft is positioning its AI infrastructure build-out as a direct response to overwhelming customer demand. The company initially projected that demand for its AI capacity would outstrip supply through the end of 2024. Now, executives have extended that forecast, expecting the supply shortage to persist until at least mid-2025.
This strategy is fortified by a massive backlog of contracted business. The company's remaining performance obligations (RPO) recently surged by 50% to nearly $400 billion. Crucially, Microsoft emphasizes that this figure represents a wide array of customers with a weighted average contract duration of just two years, suggesting this backlog will convert to revenue relatively quickly.
Microsoft's Surging Backlog
The company's RPO, a measure of future contracted revenue, jumped to almost $400 billion. This is before counting a separate $250 billion deal announced with OpenAI, indicating a powerful stream of future income tied to its AI services.
Meta's Speculative Vision
In contrast, Meta's investment strategy appears to be built more on faith than on existing contracts. When pressed by analysts to clarify the purpose of his company's escalating capex, CEO Mark Zuckerberg remained intentionally broad.
"Building new services that reach billions of people is a huge muscle that Meta has developed," Zuckerberg stated, without specifying what those future AI services might entail.
This approach asks investors to trust in Meta's track record of building massive platforms. While the company's stock has recovered significantly from its metaverse-driven lows, a recent 10% drop following the spending announcement suggests that investor patience may have its limits. The bet is that Meta can once again create a category-defining product, but the path to monetizing this massive AI investment remains unclear.
Profitability and Rising Costs Emerge as Key Concerns
As data centers fill with powerful but short-lived AI servers, a new financial pressure is mounting: depreciation. The cost of these assets is written off over their useful life, and these charges are beginning to weigh on profit margins across the sector.
What is Depreciation?
Depreciation is an accounting method used to allocate the cost of a tangible asset over its useful life. For tech companies, the rapid obsolescence of AI hardware means servers and chips purchased for billions today will be worth much less in just a few years, and that loss in value must be accounted for on the balance sheet, impacting profitability.
At Microsoft, depreciation and amortization expenses have already jumped significantly, rising from 11.3% of revenue a year ago to 16.8% in the latest quarter. This trend highlights the growing importance of operational efficiency and pricing power. Even in a boom, managing costs is becoming critical.
The impact on profitability varies. Google and Microsoft both posted stronger-than-expected profit margins recently. Conversely, Meta's operating margin fell by three percentage points to 40%. This divergence shows that simply spending on AI is not enough; companies must also build cost-effective infrastructure and command prices that justify the immense upfront investment.
The Quality of Revenue Comes Under Scrutiny
Not all future revenue is created equal, a lesson investors are learning as they dig deeper into corporate filings. The headline numbers for RPO can be misleading if not properly contextualized.
For instance, Oracle's stock soared after it revealed its RPO had jumped to $455 billion. However, it later became clear that $300 billion of that total was tied to a single five-year deal with OpenAI. This commitment is contingent on OpenAI generating enough demand and cash flow to utilize and pay for that capacity, making it a more speculative form of future revenue.
This contrasts sharply with Microsoft's more diversified and shorter-duration contract base. The distinction is crucial: one represents a broad market embracing AI services now, while the other represents a massive bet on a single, albeit important, partner.
Strategic Positioning for the Long Haul
As the initial euphoria around AI matures, the strategic decisions being made today will determine the long-term winners. Companies are not just competing on the size of their AI models but on the entire technology stack.
- Google is betting on its vertically integrated approach, developing everything from its own custom AI chips (TPUs) to its foundational models. This strategy aims for maximum efficiency and control over its AI ecosystem.
- Microsoft appears to be strategically avoiding the low-margin business of simply renting out raw GPU power. CEO Satya Nadella has hinted at focusing on more profitable, value-added services built on top of the core infrastructure.
The loosening of Microsoft's exclusive ties with OpenAI has caused some concern, but the company's leadership seems confident that a broader market of opportunities will yield greater returns. When the current AI investment party eventually slows, the companies with the most defensible profit margins, sustainable demand, and efficient infrastructure will be best positioned to lead the next phase of technological innovation.