What You’ll Find This Week

bp shut down its venture capital arm in July, citing "disciplined capital allocation," three words that explain nothing. Munich Re closed its own fund in October, the same quarter it posted a record €2.1 billion profit. Seven corporate venture funds have closed in about a year, for seven different reasons, and not one of them had anything to do with how the fund's own investments performed.

This week: why bp kept its fund running for eleven years after its own patents started declining and closed it anyway, why Commerzbank shut down one venture unit and reinforced the other in the same month, and why CVC dollars just hit a record high while corporate investors are showing up to fewer deals than they have in a decade.

Here’s what you’ll find:

Don’t Miss Our Latest Podcast

This Week’s Article

bp shut down its venture capital arm in July. Its own explanation was "disciplined capital allocation," three words vague enough to justify almost any decision a company makes. PayPal closed its fund a month prior and was more specific: a company-wide push to cut $1.5 billion in costs.

Munich Re closed its own fund in October, and its reason had nothing to do with money at all. It called the move a "strategic shift to source innovation solely from its core businesses." The company wasn't in trouble either. That quarter, it posted a record €2.1 billion profit.

Seven corporate venture funds have closed in about a year. The four that disclosed a number account for at least $4.8 billion combined investment. Read as a group, the obvious story is corporate venture pulling back under pressure. But the actual data says the opposite: CVC dollars just hit a record high. Something else is happening, and it isn't subtle once you see it.

bp Ventures ran for twenty years and deployed roughly $1.2 billion across at least 27 portfolio companies. As of today, the portfolio holding those stakes is worth about what bp put into it, an outcome close to breakeven after two decades of investing.

Munich Re Ventures made roughly a hundred investments over its ten years, funded by $1.2 billion of committed capital, including Hippo and Next Insurance. The latter bought outright by Munich Re itself for $2.6 billion.)

In ten years, PayPal Ventures backed more than 80 companies, including Plaid and Anchorage Digital, deploying roughly $850 million. Its June 2026 closure came four months into new CEO Enrique Lores's turnaround plan, which targets $1.5 billion in savings and a 20% headcount reduction over the next two to three years. A PayPal spokesperson called it part of "continued efforts to sharpen our focus," and said the company was still exploring strategic options for the unit.

None of the three verdicts said anything about how the fund's own investments had performed. bp's reason was a company-wide capital reset. PayPal's was a company-wide cost cut. Munich Re's was a deliberate reorganization, folding venture investing into its asset manager MEAG instead of running it as a stand-alone unit, but even that decision skipped the fund's own track record entirely.

The pattern holds beyond these three. ING Ventures halted new investments in May 2025 after deploying €1.5 billion since 2017, blamed on a weak exit market for its own portfolio's returns. ANZ shut its 1835i unit in October 2025, not over the fund's results but as collateral damage from a $158 million regulatory penalty, about A$240 million, and a 3,500-person restructuring that swept up all fourteen of 1835i's staff. Hypertherm paused its fund in July 2025 and never explained why to anyone who asked.

Seven funds. Seven different triggers: a capital reset, a leadership handoff, a cost-cutting mandate, a regulatory fine, a weak market, nothing at all. One outcome.

Commerzbank closed neosfer in June 2026, the innovation unit it built thirteen years earlier to help modernize the bank. The closure came under Momentum 2030, an efficiency program whose stated goal is to make Commerzbank "more modern, more efficient, and more competitive."

The bank shut down the unit built to make it modern, in the name of a program built to make it modern.

CommerzVentures, Commerzbank's other venture arm, wasn't touched. The bank explicitly reaffirmed its commitment to the team, run for over a decade by managing partners Paul Morgenthaler and Patrick Meisberger, focused specifically on fintech.

Same bank. Same efficiency drive. One venture unit closed. The other one didn't move.

If the crisis alone explained these closures, both funds would have died. Momentum 2030 hit every corner of Commerzbank's balance sheet. Whatever decided which fund survived, it wasn't the size of the emergency.

One academic finding complicates the seven-funds story. Song Ma's "The Life Cycle of Corporate Venture Capital" studied CVC programs at public US companies and found that termination often follows a recovery in the parent's own innovation output, measured by patent filings, while the fund is still active. Read that way, closing a fund isn't a failure so much as evidence the fund already did its job. Neither bp nor PayPal was in Ma's sample, which only covers firms with entries between 1980 and 2006, so the theory has never actually been tested against this specific wave of closures. Here's what happens when it is.

bp's patent filings, the standard economist's proxy for internal innovation output rather than innovation itself, peaked in 2015 at 622 and fell nearly every year after. Its R&D spend peaked in 2014 at $3.6 billion and fell to $570 million by 2025. Ma's theory predicts the opposite of what happened here: a decade of declining internal innovation should have kept the fund alive to cover the gap. Instead bp ran the fund for another eleven years and closed it anyway.

PayPal's numbers run the other direction, and break the theory a different way. Its R&D spend climbed almost every year, from $834 million in 2016, the year the fund launched, to $3.1 billion by 2025. Internal innovation never dipped enough to need the fund's help. Ma's theory predicts the fund should have kept running. PayPal shut it down anyway.

By the best proxy available, performance predicted neither survival nor death. Whatever's sorting these funds, it isn't performance. Something else is doing the sorting, and the next place to look is why these funds existed in the first place.

Dushnitsky and Lenox studied 1,173 US public firms across the 1990s and found a split:

Standard VC's entire model is betting across a wide portfolio and expecting most positions to fail while a few pay for the rest. That's the job, and specialist firms have built two decades of infrastructure, deal flow, and expertise around doing it well. A corporate fund copying that model without any of that infrastructure was, statistically, indistinguishable from doing nothing. The coefficient on strategically motivated CVC was positive and significant. The coefficient on financially motivated CVC was negative and not significant.

That finding is twenty years old. 40% of corporate venture units under $100 million in assets still run on a pure financial-return mandate today. Units founded since 2020 skew the other way. 63% cite preparing for future disruptions as a top priority, against 53% for financial returns. CommerzVentures buys fintech access for a bank. Sinopec's new $690 million fund buys a hydrogen supply chain. Toyota and Mitsubishi's funds buy the same kind of thing. Nvidia and Microsoft's AI-era funds buy chip and cloud commitments the same way. None of them is a diversified bet on the venture asset class. All of them buy something the parent needs.

The same split explains the mechanism I called murky last week: Nvidia, Microsoft, and others writing venture checks that come back as chip orders and cloud contracts. That piece's concern was pricing. A check that's really a supply contract doesn't get valued the way a normal funding round does, so a startup can't easily tell how much of the number on the term sheet is investment and how much is future revenue the parent is guaranteeing itself. Here, the same substitution explains why a fund survives a budget crisis instead. A fund that's visibly buying the parent something real is harder to cut than one that's just chasing returns nobody's tracking closely. The corporation comes out ahead here, with no real downside on its side of the table. For the startup taking the check, it's more mixed: real commercial value for as long as the relationship holds, plus a risk that has nothing to do with the startup's own performance and never shows up in the term sheet.

CVCs participated in 21.1% of US venture deals in the first half of 2026, a decade low. They accounted for 82.6% of total US VC deal value, a record high. Fewer corporate investors are showing up to rounds. The rounds they do show up to are enormous. Once buying access, not buying a return, is the real sorting test, that split stops looking like a paradox. The diffuse, many-small-checks funds, the ones built to look like a real VC portfolio, are the ones dying. The concentrated, operationally-tied funds, the ones writing checks that double as supply or customer agreements, are the ones growing. That's what produces a falling deal count next to a rising dollar figure.

Two separate things are happening here, and they're easy to mix up. The sorting rule itself isn't new. Every CVC downturn since the 1960s has killed the financially motivated cohort first: true after the 1973 slump, true after the 1987 crash, when corporate venture program counts fell by a third within five years, true again after the dot-com bust. The purchase-order mechanism isn't new either. This newsletter already traced it to 1999, when telecom vendors financed their own customers' equipment purchases and ate the loss when those customers couldn't pay. Lucent alone booked $2.2 billion in bad debt in a single year. What's new is the scale, and the divergence it produces. In 1973, 1987, and 2001, deal count and dollar volume fell together, because the checks getting cut were all roughly the same size. This time the checks aren't the same size. A handful of them, Nvidia, Microsoft, Google, and Amazon writing checks that come back as chip orders and cloud contracts in the $217 Billion Purchase Order, are large enough to keep the dollar total climbing even as the number of deals keeps falling.

A reinsurer, an oil major, three banks, and an industrial cutting-tools company just proved the underlying rule works the same way without a single AI dollar in the room. The AI checks just made an old pattern big enough to notice.

When a fund closes, the equity moves. bp is selling more than ten of its remaining stakes to Verdane, a Nordic private equity firm. Munich Re's holdings are moving to its own asset manager, MEAG. PayPal hired Jefferies to find buyers for its portfolio on the secondary market. Whoever buys those stakes gets the ownership. They don't get the reason a corporate check was worth more than a normal VC's money to the startup in the first place: the warm introduction inside the parent, the pilot program, the reference customer.

Fidelity International's venture unit shows what a clean handoff actually looks like. It sold to 7Ridge in May 2026 with the entire investment team moving alongside the portfolio. The relationships transferred because the people did, not because the paperwork did.

There's a cost running the other direction too, one that doesn't show up on any single fund's balance sheet. Cabral, Kumar, and Park studied 3,109 startups across 10,406 funding rounds and found that other investors avoid syndicating with CVCs whose past behavior was "uncommitted or capricious." A corporate investor's presence in a round usually signals that someone vetted the company. A fund that proves unreliable poisons that signal retroactively, for every founder it ever backed, and it hits hardest at the earliest-stage companies who most need the signal to mean something.

None of this is illegal or against any rule. Closing a venture fund is ordinary capital allocation, and nobody in this piece did anything they weren't entitled to do. But that doesn't make the funds that stay open automatically safer for the founders inside them. The operationally-tied funds, the ones buying access instead of returns, carry a version of the same risk: the parent funding them is also, often, their customer or their supplier or their channel. If the parent's priorities shift, the startup doesn't just lose an investor. It loses the customer relationship at the same time, from the same decision, for the same reason.

The founder is the one carrying a risk they never signed up for. The parent walks away clean either way.

What decided each fund's fate was never really in question by the time the closure hit. A fund built to make a diversified, financially-motivated bet on the venture asset class had nothing else tying it to the business, so it went the moment the parent needed cash. A fund built to buy access, a supply relationship, a pilot customer, a technology pipeline, survived the same crisis because cutting it would have cost the parent something beyond the fund itself. Most of these units were never really running as venture capital. They were running as corporate development, structured to look like a fund.

If you're the founder on the other side of one of these checks, the fund that looks safest, tied tightly to a product line, clearly not going anywhere, is exactly the one whose survival depends on your parent's priorities staying stable rather than on your own performance. Before signing, ask which bet you're actually making.