Big tech is ramping up investment in AI, but has not simultaneously acquired startups: the changing M&A landscape in 2026

Big Tech is ramping up AI investment without simultaneously buying startups: changes in the 2026 M&A landscape

Industry Context

PitchBook’s core conclusion in *Q2 2026 Building, Backing, and Buying AI* is that the AI industry is entering a phase of “heavy building, light buying.” Google, Microsoft, Amazon, and Meta are expected to spend nearly $600 billion combined on AI infrastructure in 2026, yet at the same time, these companies’ AI M&A activity remains relatively low.

This does not mean AI M&A has lost its significance; rather, it means the capital allocation logic of Big Tech has changed. Instead of acquiring startups to gain capabilities, they are more inclined to direct funds toward data centers, compute power, cloud infrastructure, and long-term strategic partnerships. PitchBook specifically notes that Microsoft’s nearly 30% stake in OpenAI and Meta’s $14.3 billion Series G investment in Scale AI are closer to “infrastructure assurance” or long-term capability lock-in than to acquisitions in the traditional sense.

There are two reasons behind this shift. First, the bottleneck in AI competition has moved from “whether the model works” to “whether compute, inference costs, and deployment capabilities are sufficient.” Second, leading platforms want to keep key capabilities within their own ecosystems to avoid the uncertainty of post-merger integration, regulatory scrutiny, and the risk of talent loss.

Market Impact

For the market, the most immediate impact of this shift is that exit paths for AI startups are narrowing.

PitchBook points out that Big Tech’s AI acquisitions fell to a 10-year low of 7 deals in 2024, rebounded to 14 in 2025, and stand at 12 year-to-date in 2026, though overall they remain below historical norms. At the same time, the VC-backed exit market has developed a “barbell” structure: on one end are a small number of highly strategic deals, and on the other are a large number of transactions with undisclosed deal values and limited returns for investors.

This means:

• For startups, simply “building a model or feature that can be acquired” is no longer enough to support an ideal exit;
• For investors, valuation logic is increasingly dependent on buyer structure, strategic fit, and scarcity of capabilities, not just product hype;
• For enterprise customers, industry leaders are more likely to stabilize supply through in-house development and long-term lock-in rather than by rapidly filling gaps through repeated acquisitions.

At the same time, capital markets are also reassessing the growth potential of AI software companies.At the same time, the capital markets are also reassessing the growth prospects of AI software companies. PitchBook notes that established enterprise software vendors such as Salesforce and Databricks are actively making acquisitions because AI agents are undermining the seat-based pricing model that defined SaaS economics over the past 20 years. The iShares Expanded Tech-Software ETF fell 29.6% between December 2025 and February 2026, reflecting that the market has begun to reprice the impact of AI on software business models.

Competitive Landscape

Who Is “Building”

Google, Microsoft, Amazon, and Meta are the core buyers in this wave of AI buildout. They are more willing to invest capital in infrastructure rather than quickly acquiring assets through M&A. The advantage of this strategy is greater controllability, and it can directly strengthen the supply of cloud, compute, and model services.

Who Is “Buying Differently”

NVIDIA is an exception. According to PitchBook, since 2022 NVIDIA has completed 15 AI-related acquisitions, a path that is clearly different from other big tech companies. Its acquisition targets are not simply about expanding market share, but about extending control across the AI full stack, creating a more complete infrastructure loop from chips and networking to software and systems.

Who Is “Forced to Buy”

Enterprise software vendors, by contrast, face a different kind of pressure. For companies like Salesforce and Databricks, AI agents are changing the way users interact with software. If users no longer buy software by “seat” but instead use AI by task, outcome, or automated workflow, traditional SaaS pricing will come under pressure. Acquiring AI capabilities has become a defensive strategy for them to respond to changes in the business model.

Who Benefits, Who Is Under Pressure

• Beneficiaries:
• AI infrastructure providers
• Cloud service providers
• Vendors that can offer inference, data, networking, and deployment capabilities
• AI startups with strategic scarcity

• Under pressure:
• AI startups relying only on functional differentiation
• Startups dependent on the “get bought by a big company” path
• Traditional seat-based software vendors

Enterprise Implications

For enterprise decision-makers, the most important takeaway from this report is not that “M&A is declining,” but that the center of gravity in AI commercialization is shifting from “acquiring capabilities” to “building infrastructure.”

Enterprises should focus on three things:

1. Is the AI supply side stable? As large companies direct more capital into infrastructure, enterprises purchasing AI services need to pay closer attention to cloud costs, inference latency, model availability, and long-term supply stability.2. Is the ROI of AI applications clear enough Against the backdrop of continued expansion in infrastructure spending, internal AI projects at companies can no longer secure budgets merely by being “technologically advanced.” They must answer how much cost will be saved, how much efficiency will be improved, and how much incremental revenue will be generated.

3. Partnerships matter more than one-off transactions In the future, companies are more likely to obtain AI capabilities through long-term cloud agreements, strategic partnerships, custom models, and co-deployment, rather than waiting for product integration after an acquisition.

For startups, capability positioning also needs to change. The closer they are to underlying infrastructure, data, inference efficiency, security and compliance, and vertical use-case implementation, the more likely they are to attract strategic investment or acquisition premiums; if they are only at the general-purpose feature layer, exit pressure will be greater.

Outlook

12 months

AI infrastructure spending will remain elevated, and leading tech companies will continue to prioritize building over acquiring. Strategic minority investments may continue to increase, especially in model-, data-, and infrastructure-related companies.

24 months

Demand for AI capability acquisitions from enterprise software vendors may rise, as the agentification trend continues to disrupt traditional SaaS revenue models. Acquisitions will be used more to fill gaps in workflows, automation, and enterprise integration capabilities.

3 years

The exit structure of the AI industry may further diverge: a small number of strategic targets will command high premiums, while more companies will need to rely on paths other than M&A to realize returns. Regulatory and geopolitical factors will also become important variables in cross-border AI transactions, and companies will assess deal structures, talent arrangements, and compliance risks more cautiously.

Conclusion

The signal conveyed by this PitchBook report is very clear: the competitive focus in the AI industry is shifting from “who bought whom” to “who can continuously build and secure infrastructure capabilities.” For big tech companies, building is more aligned with the current logic of resource allocation than acquiring; for startups, the exit window has not closed, but the threshold is higher and the paths are narrower; for enterprise customers and investors, judging AI value in the future cannot rely only on models and products, but must also look at the compute power, inference costs, business model, and regulatory structure behind them.

Source URL

https://pitchbook.com/news/reports/q2-2026-building-backing-and-buying-ai