M&A & Fundraising Analysis for October 24, 2026

Bolt Biotherapeutics absorbs Prana Therapies: a shell company reinvents itself

Source: stocktitan.net →

Bolt Biotherapeutics is a Nasdaq-listed company whose original pipeline focused on immunoconjugates for oncology. It is acquiring Prana Therapies via an all-stock reverse merger — a transaction signed and closed simultaneously — and raising $120 million in a concurrent private placement, with warrants that could generate an additional $60 million.

Prana is developing PRA-216, an extended half-life multispecific antibody, currently in Phase 2 for asthma, COPD, and atopic dermatitis, with a fourth trial planned for eosinophilic esophagitis. The underlying platform, INSPIRE, is designed to generate proprietary multispecific antibodies. Initial clinical results from all four trials are expected starting in Q2 2027. The funding is sized to cover operations through 2028. Sheila Gujrathi, co-founder of Prana, will assume strategic leadership of the combined entity and join the board.

The highlight of the structure: Bolt's historical shareholders will hold only approximately 6.9% of the capital on a fully diluted basis. This is less an acquisition than a complete recapitalization under the guise of a merger — Bolt provides the listed shell, Prana provides the program and team, and new investors provide the cash. This arrangement — a reverse merger into a listed shell to avoid the delays and costs of a traditional IPO — is an increasingly common route to public markets for mid-stage clinical assets that need liquidity quickly.

LSEG acquires Octus: when credit data becomes a fortress

Source: boursorama.com → · Sector Data & Analytics — 📬 subscribe to the Data & Analytics newsletter

Octus collects and structures data on private debt and corporate restructurings: granular information, from primary sources, in inherently opaque markets.

The immediate takeaway: LSEG expands its data portfolio, strengthens its recurring revenue, and responds to Elliott Management — the activist shareholder that had been weighing on the stock for several months. The share price jumped 6% on the day of the announcement, erasing year-to-date losses.

But the rationale stated by CEO David Schwimmer is worth a closer look. He claims that Octus provides access to highly specialized proprietary data that no large language model would be "able to retrieve anywhere." This is not a communication formula: it is a precise investment thesis. Large AI models learn from what is public, abundant, and structured. Data on private debt — unlisted issuers, confidential restructurings, early distress signals — is rare, fragmented, and its collection relies on human relationships and institutional access that no one can scrape. By acquiring Octus, LSEG is not defending itself against AI: it is precisely buying what AI cannot produce — scarcity as a bulwark.

For a French financial player — bank, asset manager, private debt fund — the lesson is concrete: the value of data no longer lies in its volume but in its irreproducibility. Generic databases are threatened; data from proprietary processes, human networks, or opaque markets become strategic assets in their own right.

HACO and Yawei: Belgian engineering under Chinese flag

Source: metaltech.be →

HACO, a Belgian manufacturer of sheet metal working machines (presses, shears, bending machines) operating under the HACO Sheet Metal and BLISSBRET Presses brands, and Jiangsu Yawei Machine Tool Co., Ltd., a Chinese machine tool manufacturer, are formalizing their strategic industrial partnership at EuroBLECH 2026 in Hanover. The agreement, for which a letter of intent was signed earlier in the year, provides for Yawei to take a majority stake in HACO.

The concrete project: HACO's site in Mouscron (Belgium) will become the European headquarters and a center of excellence for the joint Yawei–HACO platform, bringing together R&D, engineering, European management, international service, and robotics/automation projects. An international Experience Center is planned there. The development plan announces about fifty jobs in an initial phase, with an ambition of 120 to 150 employees by 2029. HACO's brands, customer relationships, and commercial identity will be preserved.

The structure of this operation follows a now established pattern in the European manufacturing industry: a Chinese player provides production scale and investment resources; the European player provides the brand, distribution network, customer proximity, and regulatory legitimacy in its home market. Mouscron as Yawei's European hub is the enclosure of continental industrial capacity through the acquisition of control of a local player — without erasing its facade.

TypeSafe AI: $870 million Series A, compute goes wild

Source: quantlogix.ai →

TypeSafe AI closes an $870 million Series A at a post-money valuation of $7.5 billion, in a round led by Andreessen Horowitz. The precise business of TypeSafe AI is not described in available sources beyond its positioning in the AI sector.

The concentration is extreme.

An $870 million Series A is a structurally unusual event: it's a seed investment that most companies never reach in their entire funding cycle. Almost all the money goes into computing infrastructure and engineer recruitment — very little into sales force or commercial deployment, which signals a bet on technical depth rather than immediate market traction. When the entry ticket for a Series A exceeds that of most European Series C rounds, it's no longer venture in the classic sense: it's infrastructure financing disguised as venture capital.

ASI: $325 million for AI that controls physical infrastructure

Source: ground.news →

ASI (formerly Airspace Intelligence) develops AI systems for the management and control of physical assets — power grids, air traffic, military operations. The Boston-based company recently secured a $99 million five-year contract with the US Army for command software, and an $875 million twelve-year contract with the FAA (Federal Aviation Administration) for air traffic management.

It raises $325 million in a Series C at a valuation of $3 billion, in a round led by Spark Capital with participation from Andreessen Horowitz and Craft Ventures. The funds will finance the expansion of research and engineering teams in Boston and large-scale deployment in the energy, logistics, and military sectors.

What distinguishes ASI from the usual flow of language model labs: its business model relies on multi-year government contracts with visible revenues, not on consumer SaaS subscriptions. AI that manages critical infrastructure — where an error costs lives or paralyzes a network — is structurally less substitutable than AI that writes emails: the barriers to entry are not algorithmic, they are regulatory, contractual, and operational. For a French institutional investor, ASI illustrates the segment of AI that least resembles a bubble: contractual revenues, sovereign clients, and markets where certification takes precedence over raw benchmark performance.

TraceWeave: $2.27 million to run AI without the cloud

Source: dealroom.co →

TraceWeave is a startup founded in October 2025 by researchers from ETH Zurich, Google, and KAIST. It is developing a proprietary post-Transformer architecture — named Phase-TITAN — designed to run AI agents on memory-constrained devices (autonomous vehicles, robots, smart glasses, drones) without cloud calls.

The technical challenge is well-documented: in standard Transformer architectures, the context cache grows with conversation length, quickly saturating the memory of an embedded device.

It raises $2.27 million in a seed round led by Mirae Asset Capital, with participation from NAVER D2SF. The company has completed a paid pilot with the autonomous driving subsidiary of a major automotive group and is validating its embedded performance with a global automotive supplier. It was a top 20 finalist in the NVIDIA Inception Grand Challenge 2026.

The positioning is deliberately radical: rather than optimizing existing models (as Nota AI, a Korean competitor, does), TraceWeave is betting on an architectural breakthrough. This is the most difficult bet — replacing the Transformer rather than compressing it — but also the only one that, if successful, creates a long-term defensible position on the model layer of embedded AI. At this stage, benchmarks are internal, and commercial validation remains embryonic: it's a bet on the foundations, not on a product.