AI Investment 2026: SoftBank’s OpenAI Bet, Meta’s Manus Deal & the Utility Pivot

The recent developments in AI investment and M&A underscore a maturing market where scale, infrastructure, and control are becoming more critical than novelty. SoftBank’s completed $40 billion investment into OpenAI confirms its transformation from venture investor to ecosystem architect. SoftBank now holds over 10 % of OpenAI and has redirected capital—including by selling its Nvidia stake—to fund infrastructure (e.g., data centers via the “Stargate” joint venture with OpenAI and Oracle) to support the next generation of AI workloads. This shift from pure model investment to compute & infrastructure mirrors a broader trend in the industry.

Meta’s acquisition of Manus signals competition focusing not only on foundational model training but also on autonomous agents capable of complex tasks. The acquisition is strategically valuable, especially given its alignment with Meta’s AI agent roadmap, but it came with non-market risk: Manus’s Chinese origin drew scrutiny, prompting Meta to buy out Chinese ownership, close China-based operations, and relocate personnel to ensure regulatory compliance and national security acceptability.

Predictions for 2026 reflect growing skepticism about AI’s taking phrase and lay emphasis on deployment, validation, and measurable impact. Experts expect:

• A transition from speculative “can AI do this?” queries to “how well, how safely, and for how many?” especially in sectors such as health, law, and enterprise applications.
• AI sovereignty to become central: more nations and enterprises will demand local control of data and model deployment. This creates opportunities for regionally rooted infrastructure providers and vendors.
• A likely shakeout in the AI startup ecosystem, especially among early-stage application layer firms without defensible margins or differentiation.

From an IB (investment banking) perspective, key deal drivers are consolidating around infrastructure (data centers, cloud/edge compute), regulatory/geopolitical risk (nationality of ownership, data sovereignty), model/IP licensing, and margins in model performance and operational cost. Valuations for foundation model leaders like OpenAI and Anthropic are very high; sustaining them will require proportionally high returns and a clear path to monetization. Meanwhile, firms lacking integration with infrastructure or being exposed to regulatory risk may face valuation compression or exit challenges.

Strategic implications include:

• Deals involving geopolitical exposure (e.g., cross-border ownership or data flows) will increasingly require governance audits, divestitures, or structural separations.
• Investors and acquirers will place more weight on infrastructure ownership and control clauses in investment agreements.
• Companies in sectors like health, law, compliance may be premium-priced due to higher barriers to entry, but also higher risk if promises don’t translate to ROI.
• Financial sponsors should prepare for later-stage funding rounds to demand predictable revenue, performance metrics, and defensible IP and compute cost advantages.

Open questions to watch:

• How will valuation multiples for foundation model companies evolve if revenue growth slows or compute costs rise sharply?
• Will infrastructure providers be able to scale without becoming unprofitable, especially in energy, real estate, and hardware supply chains?
• Can AI sovereignty efforts succeed technically and economically outside U.S. regulatory/export regimes?
• How will public policy–driven risk (e.g., AI safety, privacy, IP rights) affect deal structuring and integration costs for cross-border AI M&A?