US M&A deals of the month: September 2026

September extended the US M&A recovery without turning it into a broad boom. Large transactions remained concentrated among buyers prepared to pay heavily for technology, manufacturing capacity, specialist expertise, operating scale, and established distribution, while activity lower down the market remained more uneven.

Global M&A value rose 15% year on year during the first eight months of 2026, according to BCG, but much of that growth came from the largest transactions. The number of deals worth at least $10bn reached 37, compared with 24 during the same period of 2025, and 27 involved a US buyer, target, or both. Transactions involving North American targets reached $1.2tn, with US targets accounting for around $1.1tn.

US activity immediately before September was broader than the headline megadeal figures alone suggest. EY-Parthenon recorded a 22% year-on-year increase in the value of US transactions worth at least $100m during the three months to the end of August, alongside a 23% increase in volume. Technology led deal value, while life sciences and aerospace and defence also attracted substantial capital.

That appetite for strategic control was already evident in August’s largest US deals, when capital flowed into insurance distribution, AI routing, healthcare infrastructure, investment management, and gas transmission. September pushed the logic further towards ownership of the underlying capabilities that determine how quickly companies can expand: developer ecosystems, aerospace castings, AI research, portfolio scale, and insurance distribution.

Nvidia pays $12.93bn for Hugging Face

Nvidia announced September’s largest newly agreed US transaction on 3 September, striking a $12.93bn deal for Hugging Face. The target’s value rests on the scale of the community and software ecosystem built around its platform rather than conventional physical infrastructure.

More than 18 million developers, researchers, and creators use Hugging Face, according to Nvidia, while more than 200,000 companies use it to discover, evaluate, customise, and deploy AI. The platform hosts more than three million models, 500,000 datasets, and one million applications. Nvidia was already a major contributor, having published more than 500 models and 250 datasets through Hugging Face.

Buying the platform extends Nvidia’s reach beyond the processors and systems used to train and run AI workloads. Hugging Face sits close to the developers choosing models, adapting them, evaluating their performance, and deciding how they should be deployed, giving Nvidia a larger position within the development process that ultimately generates demand for computing infrastructure.

The acquisition also creates a delicate ownership challenge. Nvidia has said Hugging Face will remain open to models, frameworks, clouds, inference providers, and computing platforms from across the industry, with Nvidia hardware not required for development or deployment. Hugging Face derives much of its utility from that neutrality, so preserving access across competing technology stacks will influence whether its community continues to treat the platform as common infrastructure rather than the distribution arm of a single supplier.

Nvidia therefore gains a large developer network and an established route into open-model development, but the commercial value of the acquisition depends partly on resisting the temptation to pull that network too closely inside its own hardware ecosystem.

GE Aerospace brings casting capacity in-house

GE Aerospace’s $11.75bn agreement to acquire Consolidated Precision Products was September’s clearest example of M&A being used to address a physical operating constraint. CPP manufactures highly engineered castings and subassemblies for commercial aerospace, defence, and other industrial customers, employing around 6,600 people across more than 20 facilities.

GE has bought from CPP for more than 15 years. Bringing the supplier inside gives the aerospace group greater control over casting capacity as demand rises across commercial engines, aftermarket services, and defence programmes, while allowing component design and manufacturing decisions to be coordinated more closely.

Capacity is particularly valuable in aerospace because qualified components and production processes cannot be added quickly. Casting complex superalloy, titanium, aluminium, magnesium, and steel parts requires specialist equipment, skills, quality systems, customer approvals, and production experience. Buying an established supplier therefore gives GE manufacturing capability that would take considerable time and capital to reproduce internally.

The financial assumptions place significant weight on what happens after completion. GE plans to finance $7bn of the purchase price with cash and the remainder through new debt. CPP is valued at approximately 18 times expected 2027 EBITDA after anticipated net synergies, compared with about 26 times before them, and GE expects the acquisition to add to adjusted earnings per share and free cash flow during its first year.

Completion is targeted for the second half of 2027, subject to regulatory approvals and other conditions. By then, the purchase will need to deliver more than additional output. The valuation assumes GE can apply its production system across CPP, improve quality and processes, and combine engineering and manufacturing more tightly enough to support the economics of the deal.

AMD spends $8.2bn on World Labs

AMD followed later in the month with an agreement to acquire World Labs for approximately $8.2bn in shares. Founded and led by Stanford professor and AI researcher Fei-Fei Li, the company gives AMD a specialist model-research team as competition across AI computing expands beyond individual processors.

World Labs develops AI models focused on spatial intelligence and the interpretation and simulation of physical environments. Those capabilities have applications across areas including robotics and simulation, where future computing requirements differ from those created by large language models alone.

For AMD, owning model expertise provides a closer view of how emerging workloads are developing and what they will demand from future processors, systems, networking, and software. Semiconductor companies have traditionally responded to computing workloads created elsewhere in the technology industry; bringing researchers closer to hardware development allows those requirements to influence product roadmaps earlier.

The transaction also gives AMD a different route into the AI software ecosystem from Nvidia’s purchase of Hugging Face. Nvidia is acquiring a large existing developer platform spanning multiple models and infrastructure providers, while AMD is acquiring a research organisation whose work can inform the architecture of its own future computing systems.

As model development expands into robotics, simulation, reasoning, and physical environments, the competition between semiconductor companies increasingly involves the research and software that determine what future hardware must be capable of doing. AMD’s purchase price reflects the value it places on gaining that expertise directly rather than relying entirely on external model developers to set the direction.

IRT and Centerspace combine at $8.1bn

Independence Realty Trust and Centerspace agreed an all-stock merger that will create a residential REIT with a pro forma enterprise value of approximately $8.1bn and equity market capitalisation of about $5bn.

The $8.1bn figure represents the value of the combined business rather than consideration paid for Centerspace. Once completed, the enlarged group is expected to own 44,354 apartments across 163 communities in 17 states. IRT shareholders will own approximately 78% of the company, with Centerspace shareholders owning about 22%.

Geographic diversification accompanies the increase in scale. Around 58% of pro forma net operating income will come from Sunbelt markets, with 27% generated in the Midwest and 15% in the Mountain West. The companies expect approximately $24m of annualised synergies and forecast that the combination will lift 2027 core funds from operations per share by about 5%.

Operating systems and capital access are central to those estimates. IRT plans to apply its renovation programme, technology, Wi-Fi revenue initiatives, and other operating practices across a larger estate, while the combined company’s greater market capitalisation and free float are expected to increase its weighting in several property and equity indices.

Unlike acquisitions whose returns rely on adding substantial new borrowing, the merger is intended to remain leverage neutral. Combining portfolios allows the companies to increase operating density, spread corporate costs across a larger asset base, and improve access to institutional capital without making additional leverage the principal source of growth.

Baldwin goes private in $7.7bn transaction

Private capital supplied another of September’s largest transactions when Sequence Holdings and Dell Family Office agreed to acquire a majority interest in The Baldwin Group in an all-cash deal carrying an enterprise value of approximately $7.7bn.

Baldwin shareholders are set to receive $32.50 per share, an 88% premium to the company’s unaffected closing price on 17 June, the day before reports emerged that it was exploring a take-private transaction. The enterprise value comprises an equity purchase price of approximately $4.6bn and around $3.1bn of net debt expected to be assumed or refinanced.

Eligible Baldwin employees will retain a significant minority interest alongside Sequence and Dell Family Office. That structure keeps part of the company’s workforce economically exposed to future performance while moving Baldwin outside the public markets and placing it under owners with a longer investment horizon.

The valuation leaves substantial work to do after completion. At approximately 20 times trailing 12-month adjusted EBITDA, the transaction requires continued growth and operational improvement to support the premium. Insurance distribution can produce attractive economics through recurring client relationships, specialist knowledge, producer networks, and opportunities to use technology across a large customer base, but those strengths depend heavily on retaining people and relationships after ownership changes.

Baldwin also extends a consolidation pattern visible in August, when Aon agreed its $17bn acquisition of USI Insurance Services. The buyers and deal structures differ, but both transactions place high values on established insurance distribution networks that would be expensive and slow to recreate organically.

Scarce capabilities continue to attract capital

September’s transactions continued a US deal recovery that remains concentrated towards the upper end of the market. Large and megadeals are running above their longer-term averages, while BCG found activity among transactions below $1bn remained weaker. Capital is available, but buyers are deploying it unevenly.

The largest September acquisitions put that capital behind assets with established capabilities rather than speculative expansion plans. Nvidia gains access to a developer community already operating at global scale, while GE secures specialised manufacturing capacity. AMD brings AI researchers closer to its hardware roadmap, IRT and Centerspace create efficiencies through portfolio scale, and Baldwin’s new owners acquire an established insurance distribution platform.

Transactions immediately below the top five followed similar economics. Flex agreed to acquire California-based EPC Power for $4.4bn, adding power conversion technology as its Cloud and Power Infrastructure business prepares to become an independent company. EPC Power designs systems for data centres and grid applications, including equipment for the transition towards 800V data centre architectures.

The company expects EPC Power to generate around $800m of revenue in 2026, with approximately 40% organic growth forecast for 2027. Flex is therefore paying for more than participation in rising data centre investment. EPC brings engineering, manufacturing, hardware, software, controls, and deployed power conversion capacity into a business already supplying power, cooling, and computing infrastructure.

AI investment runs through several of these transactions, although its effect is broader than direct acquisitions of AI companies. BCG found that the technology is creating new targets while also making some existing businesses harder to value as buyers assess how durable their products and competitive positions will remain. Nvidia and AMD are buying directly into model and developer ecosystems, while Flex is investing in the electrical infrastructure required to support higher computing densities.

PwC’s 2026 M&A Integration Survey found that only about one in three acquirers fully achieved their stated deal objectives, while companies that performed better were more likely to define important operating decisions before completion and assign clear ownership to value initiatives. September’s largest deals put unusually large sums behind assets whose value depends on maintaining technical capability, customer relationships, operating discipline, or ecosystem trust after ownership changes.

September did not produce an indiscriminate rush back into US M&A. It produced a series of substantial commitments by companies and investors willing to pay for technology, capacity, expertise, networks, and scale they judged difficult to reproduce quickly. With large deals continuing to account for a disproportionate share of the market recovery, the strongest transactions are being built around specific operating advantages and a clear plan for what ownership allows the buyer to do differently.