Two Great Reads

This post is to flag two great posts by authors always worth reading, both related to the sad state of the venture capital industry upon which I have pontificated several times:

• Molly White's The venture capitalist's dilemma.
• Fais Khan's Zero Knowledge Influencer: Are ZKPs Worth the Hype?.

Each will reward your time. Below the fold I comment on both of them.

Molly White

White follows the "inverted pyramid structure by placing the TL;DR in her first paragraph:

A little less than a year ago I made some venture capitalists very angry when I made an offhand remark in an episode of Crypto Critics’ Corner: “I mean, I would probably argue that venture capitalists are not good for society regardless of what they’re investing in.” I am always surprised at how controversial a statement that is, and how much it attracts the kind of “not all venture capitalists!” sort of reaction that’s usually reserved for criticism of cops and landlords.

The occasion for her post was the unedifying spectacle of VCs triggering a run on Silicon Valley Bank then clamoring for a bail-out:

We have found ourselves in a scenario where the investor class has, yet again, managed to privatize profits and socialize losses. While many of these powerful, wealthy, and connected individuals have pushed for policies that would scale back government and regulators, promoted cryptocurrencies they believe to be outside control of the state, and pushed back against any action to break up tech monopolies, they quickly found themselves begging government officials for a rescue. “No atheists in a foxhole. No libertarians in a bank run,” tweeted Eric Newcomer, after right- and libertarian-leaning David Sacks tweeted at government officials demanding they “Stop this crisis NOW”.

As usual, White's strength is focus on the big picture:

People are realizing that despite the hundreds of billions of dollars being deployed each year by venture capital firms in pursuit of “innovation”, the world doesn’t really feel hundreds of billions of dollars better off for it. For all the talk of unbridled innovation, venture capital services only very specific types of innovation: those that stand to produce large exits for investors, and with relatively low risk, regardless of whether the business itself holds much promise or provides any societal benefit.

She quotes Edward Ongweso Jr.:

For the past 10 years venture capitalists have had near-perfect laboratory conditions to create a lot of money and make the world a much better place. And yet, some of their proudest accomplishments that have attracted some of the most eye-watering sums have been: 1) chasing the dream of zeroing out labor costs while monopolizing a sector to charge the highest price possible (A.I. and the gig economy); 2) creating infrastructure for speculating on digital assets that will be used to commodify more and more of our daily lives (cryptocurrency and the metaverse); and 3) militarizing public space, or helping bolster police and military operations.

The focus on the societal externalities of today's VCs contrasts with my narrower focus on the fact that, unlike in the good old days when I did VC-funded startups, what they are doing today isn't good for their investors or for the companies they fund. Starting almost two years ago, I discussed this in:

• Venture Capital Isn't Working
• Venture Capital Isn't Working: Addendum
• Counterpoint On Venture Capital

Source

 One key graph is this one, from a 2020 Morgan Stanley report. It shows that the media VC return was greater than the median public market return only from 1987 to 1997, and that the weighted average VC return was greater than the weighted average public market return only from 1988 to 1999 (not coincidentally when I was doing my three startups!). What the VCs have been doing since hasn't rewarded their investors for the added risk they take.

Source

It isn't just that the VCs harm their investors, they also harm the companies they fund. Another key graph, from Jeffrey Funk's The Crisis of Venture Capital: Fixing America’s Broken Start-Up System, shows that the vast majority of even the vaunted "unicorns" can't make a profit. In the good old days a company had to be profitable before it could IPO. Sun Microsystems had nearly 4 years of profits before its IPO. Now, most aren't profitable long after they IPO but they still get a billion-dollar valuation.

Clearly, from the 80s to the mid-90s VCs were doing something right. They built a slew of great companies, provided huge numbers of well-paying jobs which, in turn, fed vibrant local economies not just in Silicon Valley but also for example in Austin TX, Pittsburgh PA and Boston MA, and generated great returns for their investors.

Source

But the last was their downfall. In the late 90s these returns attracted far too much money into VC funds, and the result was the irrational exuberance that led to the dot-com crash:

Between 1995 and its peak in March 2000, the Nasdaq Composite stock market index rose 800%, only to fall 78% from its peak by October 2002, giving up all its gains during the bubble.

The more money that goes into VC funds, the lower the returns, because the worse the companies that get funded, because the less due diligence the VCs do before investing (they just watch Sam Bankman-Fried play League of Legends), and the less non-monetary help they provide afterwards.

Source

The low-interest-rate environment after the Global Financial Crisis led to the same problem that caused the dot-com crash, a flood of money chasing the non-existent above-market returns of VC funds. And, as in the late 90s, this led the VCs to fund a lot more, a lot worse companies. The previous graph goes to 2017 and shows the problem ramping up, whereas this one starts in 2016 and shows the VCs lavishing $30B on around 1,400 early-stage deals in Q4 2021 alone, or an average of over $21M. The quarter before they had lavished $60B on about 1700 late-stage deals, or an average of over $35M. Doing something like 1,000 deals a month doesn't leave much time for the due diligence and hand-holding that are supposed to be the source of the VCs' supposed above-market returns.

Instead what we get are herds of VCs chasing after the "next big thing" such as NFTs, Web3, and now AI. This isn't a new phenomenon either. Even if the "next big thing" turns out to have legs, unlike NFTs and Web3, Brian Arthur teaches that there will only be a very few big winners because of technology's strong economies of scale, which provides a huge first-mover advantage.

Nvidia is an example of how to do it right. We started in a downturn, being the only hardware company to be funded in that quarter. We could thus get the attention of the best VCs, and the best lawyers, and we could get cheap office space. Six months later the economy looked much better and we knew over 30 other companies trying to do the same thing. We had a six-month start on them, one entire product cycle of the PC industry at that time, and only ATI out of the 30 survived. It was eventually acquired, leaving Nvidia as the sole survivor.

At some point the VC industry has to recognize the fact that they paid way too much for lots of really crappy companies. It looks like Silicon Valley Bank may have triggered the moment when reality strikes, as Diana Li reports from a conversation with Bloomberg Intelligence's Gaurav Patankar in SVB Collapse Could Mean a $500 Billion Venture Capital ‘Haircut’:

The $2 trillion venture capital industry could see portfolio markdowns of 25% to 30% — a “haircut” of possibly $500 billion — following the Silicon Valley Bank debacle, according to Bloomberg Intelligence.
...
Some VC and private equity firms are turning toward strategies to “extend” and “pretend,” meaning they would hold on to assets or prop up capital to avoid true price discovery,
...
But those methods can only delay but not deny the ultimate fundamental problem of “untenable” and “unrealistic nature” of venture valuations, he said. “There are enough zombie companies with frothy valuations that need restructuring, price discovery and of course re-tooling of their business models to a world of tighter credit, subdued revenue and higher rates,” Patankar said.

'tis a consummation Devoutly to be wish'd.

Fais Khan

The poster child for today's VCs is Andreesen Horowitz (a16z), the "the SoftBank of crypto". Khan's subhead is "going after a16z yet again", a reference to two previous (excellent) posts:

• "You Don't Own Web3": A Coinbase Curse and How VCs Sell Crypto to Retail (see here).
• The Unstoppable Grift: How Coinbase and Binance Helped Turned Web3 into Venture3 (spoiler: by trading startups' coins):
  

  [VCs are] collectively shoveling what will probably amount to $10B this year into the crypto startup market - enough for 300 $30m Series A’s. Honestly, I had to laugh writing that. Like a poor French goose, whatever half-decent startups might be out there in crypto land are probably being hounded by hungry investors, while the market gets drowned in a flood of new coins.
  
  With hundreds of millions in liquidity available less than one year from investment, that creates a snowball effect. Trading revenues turn into VC rounds that turn into more trading revenues. How else has a16z gone from a $300M fund to $2.5B to $4.5B in less than four years?

David Gerard's Web3: a VC-funded gig economy of securities fraud explains the "unstoppable grift":

The entire venture capital push for Web3 is so that Andreesen Horowitz (a16z) and friends can dump ill-regulated tokens on retail as fast as possible. This gives the VCs very fast liquidity events — much faster than they get from investing in actual companies.

The VCs try not to commit the securities fraud themselves. Their business model here is to incentivise it.

Cashing out from a startup can take years — but dumping a minor altcoin DAO governance token on retail can be completed in just a few months."

And Ed Zitron summarizes:

it’s just a system through which venture capital can quickly see ridiculous returns while promising (and letting down!) those who are simply buying a token because they believe it’ll be worth more because of its presence on Coinbase.

Khan starts:

New year, same old me: taking shots at crypto narratives.

And crypto is searching for a new one after 2022’s bloodbath. So far in 2023, a16z has shown us the way with a $100m round for Aztec, a zero-knowledge proof focused company my old employer ConsenSys was one of the first investors in (disclosure, I wrote a memoir about ConsenSys that you should check out).

Wikipedia's zero-knowledge proof examples are a good way to understand the basic idea — a verifier can repeatedly challenge a prover with a problem that, if the prover knows the secret, they will answer correctly 100% of the time, but that if they don't, their random guess will be right 50% of the time.

Khan's skepticism is based on experience:

ZKPs are being touted as the next catalyst for crypto, a “game changer” that will make blockchain safe and scalable using revolutionary technology. It seems odd I would choose to criticize something so promising, but ever since I first worked with a ZK team five years ago I found one fact troubling: that ZKPs have actually been around for 3 decades but never found much adoption - although some of the related cryptography proved to be important - a story that seemed all too familiar to me. I started to grow cynical.

One of the fundamental problems with ZKPs are that the cryptography needed is computationally expensive. Thus there are three design choices for a system using it:

• Accept the cost by making ZKP-based privacy mandatory (Monero). Khan calls this "fully private".
• Allow users who need it to pay the cost by making it optional (Zcash). Khan calls this "partially private".
• Amortize the cost across a batch of transactions. Ethereum's future "rollups" take this approach.

Khan explains:

There’s a “ZKP trilemma” similar to Vitalik Buterin’s “blockchain trilemma”: no solution can offer untraceable privacy, compliance with the law, and decentralization. When it comes down to it, the only reason to use a ZKP today on a blockchain is to commit a crime.

Source

Khan comments on the "fully private" option:

How does Monero handle these issues? You simply declare that regulations don’t apply to you, and become the #2 preferred currency of the dark web!

The "partially private" option has the problem that, in practice, people don't want to pay the considerable extra cost for privacy:

Zcash requires massive computing power and user skill, and even then these ZKP-based blockchains can only do the simplest of operations - transfers. The time and cost of these remains slow - often more than 30 seconds, and 10x+ the cost of an already expensive Ethereum transaction. Years of research on specialized hardware likely remain ahead.

Source

The result of the choice and the exorbitant cost is obvious:

In fact, we can see this in Zcash’s own website, which shows less than 20% of transactions are at all private, and Chainalysis notes that less than 1% are “fully” private. This number has more or less held steady for years, except for a spike around a well publicized roll out. And worth noting, we have no idea if Zcash is fudging those numbers by entering its own transactions!

The other problem with making privacy optional is that users who can chose not to pay may understand that the privacy isn't very good:

Numerous studies show that blockchain analytics firms like Chainalysis can trace transactions due to the need to “convert” money from public to private and back, especially when exchanges like Coinbase don’t allow shielded withdrawals.

Users can also take the fully private option and use Monero instead. This might be why Zcash has a "market cap" around $570M and daily volume around $23M, whereas Monero's "market cap" is around $3B, and daily volume around $106M, so Monero is 5 times larger. Monero's use in the Dark Web is around 25 times that of Zcash, so it is obvious that users for whom privacy is important know where to get it and are willing to pay for it. This is because the penalties for crime-ing with inadequate opsec are severe.

Khan summarizes the other big ZKP problem:

All of the excitement around “web3” is for smart contracts, but there are no functional zero knowledge smart contract chains. This is quite limiting - without it, there are no DeFi applications, loans, yield, and all the other things people have been excited about. While there are some theoretical and testnet-type implementations, they have significant caveats. That’s because most of the ZK implementations so far use the UTXO model, similar to Bitcoin, which significantly impairs programmability.

Do "rollups" solve the "ZKP trilemma"?. First, privacy:

Zkrollups currently require you to lock up assets on Ethereum, which means your assets are still transparent. The blockchain is then updated in batches, which means you can likely trace transactions and figure out counterparties from within the batch. Unless you use a mixer…and you know where that will go.

Add to that that none of the zk-rollups today are actually private, and so this is just theoretical/promises today.

Second, decentralization:

Zk-rollups require a sequencer, which means a centralized party that puts transactions in order. This is the key to achieving “Visa”-like scalability - but seems to defeat the entire purpose of using a blockchain in the first place. How long would it take until a zk-rollup team rug pulls by issuing an infinite amount of tokens - with no one even knowing about it?

Most rollups are custom built (and sometimes run) by centralized companies, like StarkWare and Polygon. They have built rollups that are not fully open source or that are custom versions of EVM, so the company needs to exist to maintain/debug the code, and traditional corporations would have to build apps specifically for these environments. Unlikely!

Third, compliance:

As centralized entities that process financial transactions, rollups would likely be under significant regulatory scrutiny under banking, AML, and securities laws. If the rollup was decentralized, as some plan to do “eventually,” these smaller chains could become targets for hacks, front running (MEV), or spam.

Thus Khan argues that all proposed uses of ZKP technology, while attention-grabbing and mysterious, suffer from the "ZKP trilemma" and are unlikely to change any games any time soon, disappointing the VCs.

Source

In particular, the compliance prong of the trilemma is gaining in importance. Recent experience shows that you can get away with non-compliance for a while but eventually among the outcomes are that you get a Wells notice, get sued by the SEC (who have a 130+ to 0 record of success in cryptocurrency cases), get hauled out of your private plane at Podgorica Airport, get jailed in the Bahamas, or turn up "dead" in India. These are probably not optimal outcomes for you. Nor for the VCs backing your ZKP startup, who these days because of the Greater Fool Supply-Chain Crisis are less likely to have taken the money and run via List And Dump Schemes.

I urge you to read the whole of Khan's post, because above I just extracted the parts that fit the story about VCs. There is a whole lot more.