When SoftBank Group announced its $4 billion definitive agreement to acquire DigitalBridge Group on December 29, 2025, the deal landed in the press as a vertical-integration story for the Stargate program. SoftBank, the financier; DigitalBridge, the data center owner; the combination, an AI infrastructure flywheel. The structure made sense and the strategic logic was clean, but in late December the deal felt comparable in significance to a dozen other large infrastructure announcements that had moved through 2025.

Five months on, that read is wrong. Three things have happened in the interval that reframe the deal entirely. Nvidia used its GTC 2026 conference in April to launch an enterprise AI agent platform with seventeen launch partners including Adobe, Salesforce, and SAP. Google's Cloud Next 2026 conference in early May positioned Vertex AI and Gemini Enterprise as the foundation for "continuous agent execution." And SAP Sapphire 2026 in mid-May announced production-ready agentic capabilities for RISE on Microsoft Azure, with Microsoft committing to more than double the number of customers in the joint acceleration program.

These announcements share a thesis: enterprise AI in 2026 is no longer a model question; it is an infrastructure question. The companies that win are the ones that own the physical and platform substrate underneath the agents. And the company best positioned to win that fight is the one that financed both the agents (OpenAI) and the substrate (DigitalBridge) — which is SoftBank.

The deal that read like a financing story in December reads like a control story in May. For enterprise CIOs and the boards above them, the difference matters.

The deal mechanics

SoftBank Group Corp. (TSE: 9984) entered into a definitive agreement on December 29, 2025, to acquire DigitalBridge Group, Inc. (NYSE: DBRG) at $16.00 per share in cash. The price represented a 15% premium to DigitalBridge's December 26 closing price and approximately 50% premium to the unaffected share price before deal rumors leaked. The transaction values DigitalBridge's equity at roughly $2.92 billion and the total enterprise value at approximately $4.0 billion.

DigitalBridge managed roughly $108 billion in digital infrastructure assets under management as of September 2025, across portfolio holdings that include Vantage Data Centers, Zayo, Switch, and AtlasEdge. The portfolio spans data centers, fiber networks, cell towers, small-cell systems, and edge infrastructure — every layer of the physical substrate that AI workloads run on, with the exception of the GPUs themselves.

The transaction is expected to close in the second half of 2026, pending regulatory approvals and customary closing conditions. SoftBank disclosed the agreement via a Form 8-K filing with the SEC and a joint press release on December 29, 2025.

What makes the deal structurally interesting is not the multiple — 15% to closing price is unremarkable for a strategic acquisition — but what SoftBank gave up to make the deal possible. In November 2025, SoftBank sold its entire remaining Nvidia stake for $5.83 billion to fund its OpenAI and Stargate commitments. Founder Masayoshi Son later said publicly that the Nvidia sale made him cry. The DigitalBridge acquisition draws on the same capital pool — SoftBank is converting liquid public-market exposure to AI compute (the Nvidia stake) into ownership of the infrastructure that runs the compute (the DigitalBridge portfolio). The bet is that owning the substrate produces more durable returns than holding equity in the chipmaker.

The Stargate connection

DigitalBridge is one of the primary backers of Stargate, the OpenAI-Oracle-MGX-SoftBank partnership announced in early 2025 to build advanced AI computing campuses in Texas, New Mexico, and Ohio. The campuses are designed to deliver up to seven gigawatts of combined power and serve as the backbone for large-scale AI training and inference workloads. Vantage Data Centers, a DigitalBridge portfolio company, is building one of the Stargate facilities — a near-gigawatt installation in Wisconsin.

Before the DigitalBridge acquisition, SoftBank's Stargate position was that of a co-investor whose returns depended on DigitalBridge executing well. After the acquisition closes, SoftBank owns the executor. The change is not cosmetic. It moves SoftBank from a position of capital exposure to a position of operational control over the largest committed AI buildout in the United States.

For enterprise customers that will eventually buy compute capacity from Stargate facilities, the consolidation is double-edged. On one hand, a single owner across capital, models (via OpenAI), and infrastructure can move faster than a coordinating consortium — fewer veto points, cleaner execution. On the other hand, single ownership concentrates pricing power and reduces optionality. CIOs buying compute commitments in 2027 and 2028 will be negotiating with one counterparty across more layers of the stack than they would have been five months ago.

The Roche announcement at GTC in April offers a useful comparison. The pharmaceutical company committed to deploying more than 3,500 Nvidia Blackwell GPUs across hybrid cloud and on-premises environments in the U.S. and Europe — what Roche calls the largest announced GPU footprint available to a pharmaceutical company. That deployment is a $1+ billion infrastructure commitment by a single enterprise customer. Multiply Roche's pattern across the Fortune 100 and the demand picture for AI infrastructure over the next 36 months is the largest committed capex cycle in enterprise technology history.

The company that owns the substrate that picture runs on is no longer just a financier. It is a utility.

What the deal does and does not buy

Three things, in order of analytical confidence.

It buys SoftBank vertical integration across the AI stack. Pre-deal, SoftBank had exposure across models (OpenAI equity), chips (Nvidia equity, now sold), and infrastructure (DigitalBridge equity). Post-deal, it owns one of those layers outright and retains structural exposure across the other two through Stargate and direct OpenAI commitments. The integration is not as clean as Nvidia's silicon-to-platform play, but it is the closest thing to it in the financier-operator quadrant.

It buys timing. The DigitalBridge transaction was priced at December 2025 multiples. Every major infrastructure announcement since — Nvidia's enterprise agent platform, Google's Vertex AI expansion, SAP Sapphire's agentic capabilities, the Roche Blackwell commitment — has implicitly raised the strategic value of owning physical AI infrastructure. SoftBank locked in the price before the strategic re-rating became consensus. If the deal had been negotiated in May 2026 instead of December 2025, the premium SoftBank paid would almost certainly be higher.

It buys financing optionality. DigitalBridge as a standalone alternative asset manager had a cost of capital constrained by public-market sentiment, sector rotation, and the broader REIT-and-infrastructure pricing environment. As a SoftBank subsidiary, the same infrastructure assets can be financed against SoftBank's balance sheet, Vision Fund capacity, and Saudi Arabia's PIF relationship. The asset side of the equation is unchanged. The liability side gets dramatically more flexible.

What the deal does not buy: pricing power over Nvidia. The chip cost remains the chip cost. SoftBank still has to buy GPUs at market rates, and Nvidia retains the pricing leverage that comes from being the only credible supplier of frontier training hardware at scale. The deal also does not solve SoftBank's Saudi exposure question, its Vision Fund mark-to-market problems, or the founder-succession risk that is now a standing line item in any SoftBank analyst note. The acquisition is a strategic upgrade, not a structural fix.

What enterprise CIOs should read in the contrast

Three operational considerations for any CIO or CTO planning AI infrastructure commitments over the next 24 months.

First, the vendor landscape is consolidating faster than enterprise procurement cycles are designed to accommodate. Five months ago, an enterprise considering large-scale AI compute commitments would have evaluated DigitalBridge-owned facilities as one option among many independent operators. Today, the same evaluation runs into SoftBank-aligned ownership across the model layer (OpenAI), the orchestration layer (Stargate consortium relationships), and the infrastructure layer (the consolidated DigitalBridge portfolio). Procurement teams that mapped the landscape in 2025 should re-map it in mid-2026 before signing any commitments longer than 36 months.

Second, multi-cloud and multi-vendor strategy is becoming more expensive and more important simultaneously. The SAP Sapphire announcement of agentic capabilities running natively on Microsoft Azure, combined with Google Cloud Next's positioning of Gemini Enterprise as a multicloud platform, signals that the hyperscalers are now actively competing for the same enterprise AI workloads that Stargate and similar consolidated infrastructure will host. CIOs that lock into a single vendor across the agent, platform, and infrastructure layers are exposed to pricing terms that get worse as the consolidating party gains share. The cost of multi-vendor optionality is rising; the cost of giving it up is rising faster.

Third, the contract structures that worked for cloud commitments in the 2017–2023 period are not adequate for AI infrastructure commitments in the 2026–2030 period. The variables are different — gigawatts of power, multi-year GPU delivery schedules, regional grid access, water rights for cooling — and the counterparty risk is concentrated in a smaller number of entities with deeper vertical integration. Procurement leaders should be involving infrastructure, ESG, and legal at the term-sheet stage rather than the contract-execution stage, because the commitments being negotiated now will outlive most CIOs' tenures at their current companies.

The SoftBank–DigitalBridge transaction will close in the second half of 2026. By then, the AI infrastructure landscape will likely have shifted further. Nvidia will have closed at least one or two of the seventeen GTC partnerships into deeper agreements. Microsoft and SAP will have measurable Sapphire customer counts. Google will have published Vertex AI enterprise adoption metrics. Stargate Phase 1 facilities will be partially operational. The DigitalBridge close will happen in a market that looks materially different from the one in which the deal was priced.

Masayoshi Son's framing of the acquisition in the December announcement was that DigitalBridge would "advance our vision to become a leading ASI platform provider." Strip the artificial-super-intelligence rhetoric and the strategic claim is simpler: SoftBank intends to own the infrastructure underneath the AI economy. In December, that claim was a financing story. By May, it has become an operating claim. And for enterprise CIOs writing procurement contracts that will define their organizations' AI cost base for the rest of the decade, an operating claim by a counterparty with this much consolidated leverage warrants closer reading than it received the first time around.

"The deal that read like a financing story in December reads like a control story in May."

Share thisFacebookXLinkedIn