Begin Capital's $100M Fund II: 9 Deep-Tech Investments and Thesis

Begin Capital announced a $100 million second fund on October 28, 2026, focused on companies solving difficult technical problems in AI infrastructure, hardware and silicon, and physical AI. It invests globally but emphasizes European technology talent. The London-based manager named nine initial investments: eight identifiable companies and a ninth in stealth.

What matters beyond the headline is the change of strategy. Begin's original fund backed a varied set of software and marketplace businesses. Fund II is intended to concentrate on infrastructure and physical-world technology while continuing to make flexible first cheques and invest further in the strongest holdings.

Begin Capital Fund II: key facts

A fund size is not the same as the amount already deployed. Nor does the sum of a portfolio company's financing rounds tell us how much Begin Capital itself invested.

All nine investments named at launch

Editorial distinction: this table records an investment association announced by Begin. It does not assert that every company completed a new funding round on October 28, or that Begin led every earlier round.

Funding history: what was actually raised?

These are startup-round totals supported by original company/founder statements or identified transaction reports. They are not the individual cheque amounts invested by Begin Capital.

For Areg.AI, co-founder and CEO Hayk Harutyunyan described a $5 million financing led by Begin Capital with Flyer One Ventures, Ratio Investments and other investors. He also described live deployments on 12 utility-scale solar plants during 2025. A September 2026 company update documents a 5.073 MWp project in Armenia. Its projected production and cost improvements are company targets, not independently measured results.

For SurrealDB, the $23 million extension increased the company's Series A total to $38 million and its reported lifetime financing to $44 million. These amounts are not three different rounds to sum together. For xmemory, the founder named Begin alongside Fly Ventures, AAL VC, 33East and Inovia Capital in the pre-seed syndicate.

Tomorrow Things is a particularly good data-quality check: the company had a €1.5 million seed in 2024 led by HTGF and neoteq. Begin Capital is listed in the 2026 portfolio but is not named as an investor in that 2024 release. The precise date or terms of Begin's own investment remain unavailable.

A ninth name in the website directory is not the undisclosed startup

The manager's official portfolio currently displays Finom with the label Season 2. However, its October 28 announcement lists eight different public names plus one undisclosed investment and does not mention Finom. Finom is a separate, identifiable fintech company. This may reflect differences between the website's editorial season labels and the announced fund's legal portfolio, but it does not justify counting Finom as the stealth investment or silently expanding Fund II from nine to ten. VCT keeps the reported launch list and flags the directory difference for verification.

Which companies have reached the physical world?

There is a meaningful difference between a database download, a software pilot, and a deployed industrial system. ALL3 says more than 100,000 square metres of residential projects passed through its design software by April 2026; that is a pipeline/processing metric, not evidence that all were robotically constructed. Areg.AI has documented a named solar-site deployment. Crewline markets retrofit kits for rollers and dozers, with dump trucks planned later; those availability claims come from the company rather than an independent field audit. xmemory emerged with design-partner pilots, while SurrealDB disclosed developer adoption and a general-availability product release.

For Fund II, investors should monitor actual active deployments, time-to-install, paid customer retention, product gross margin, safety/uptime evidence and follow-on ownership, not collapse diverse technology categories into one generic 'AI startup' metric.

What has actually closed? The $100 million question

Begin's October announcement described a $100 million Fund II. A September 29, 2026 US SEC Form D filing for Begin Capital II, LP, which Dealroom connects to the fund, reported an offering amount of $100 million and approximately $39.7 million sold to 12 investors at that earlier filing date.

This is not necessarily a contradiction. A public announcement can describe a vehicle's target or total reported commitments, while a filing records securities sold as of a specific earlier date. The September filing does not establish the final amount committed as of October 28 or November 9.

For a research database, treat announced fund size, securities sold at a dated filing, and independently verified final close as three separate facts. It would be premature to state that $100 million of investor money had already arrived or had been invested into startups.

From a $20 million first fund to deep tech

Begin's original investment programme started in 2020. The firm reports 28 Fund I investments, including Runware, Codesphere, Dwelly, Native Teams and Fundraise Up.

Co-founder Alex Menn says Begin started Fund I with $20 million, later increased the vehicle to $72 million, and achieved a 22% hit rate. He also describes Fund I as ranking in the top 1% of its vintage.

These performance figures are manager-reported claims. No complete comparable benchmark, net IRR, DPI, TVPI, audited cash-flow history or definition of a successful investment was established by the public sources reviewed. They should not be represented as independently proven returns.

The original fund helped Begin build relationships with operators. The new vehicle aims to use these relationships to source companies before the larger financing rounds that attract more established venture investors.

Portfolio founders become venture partners

Begin announced that Flaviu Radulescu of Runware, a Codesphere founder and CEO, and Dan Lifshits of Dwelly had joined as venture partners. It also said several original portfolio founders invested in Fund II as limited partners.

This is a potentially valuable sourcing strategy. People who have shipped developer platforms, built AI infrastructure or scaled operational businesses may evaluate technical teams differently from generalist investment professionals. However, it is too early to conclude that the model improves financial returns.

Codesphere identity resolved: The new-fund announcement names Elias Schneider, while Begin's team page uses Elias Groll. Crunchbase explicitly records Elias Groll as an alternative name for Elias Schneider, and Codesphere's June 2026 imprint lists Elias Schneider as managing director. The names describe the same co-founder; VCT does not create two person records.

How the fund plans to invest

Begin's Season 2 website sets out a concept of roughly 28 investments, with an illustrative budget of approximately $3 million per company and flexibility from $500,000 up to $8 million.

These figures describe an intention, not 28 already completed deals and not a fixed capital allocation. A $100 million vehicle funding physical AI must consider reserves, later-round participation, fund fees and the possibly greater funding needs of hardware companies. Dividing headline fund size by a target company count would not capture those constraints.

The issue to monitor is ownership retention. A manager may discover excellent robotics or semiconductor teams at pre-seed but be diluted substantially when capital-intensive manufacturing and commercial deployment demand larger later rounds.

How to verify Fund II investment claims

1. Check the manager's official announcement for fund size and the named investees.
2. Compare startup funding disclosures, for example All3's seed release or SurrealDB's Series A extension, without treating the round total as Begin Capital's own cheque.
3. Keep the vehicle's reported size separate from securities sold in the earlier SEC filing.
4. Leave undisclosed deal terms, ownership stakes and investor returns unknown until independently verified.

What investors and founders should watch

Four developments would offer stronger evidence than the announcement alone: a verified fund final close; company-level rounds that explicitly name Begin as lead or participant; measurable deployment and commercial adoption for the robotics and industrial companies; and follow-on investment by Begin during later fundraisings.

The announcement is a clear strategic statement. The eventual results will depend on whether the firm's early-check and follow-on method transfers from software to deep-tech businesses with very different technical and financial risks.

Research methodology: Fund and cheque figures are attributed to the manager unless verified in a filing. Company financing values describe rounds raised by the businesses, not individual investor contributions. Financial performance claims are not treated as verified realised returns.