Ep 18 - On The Evolution of VC in Web3

Bullish:

• $ETH - Tomasz Tunguz (Theory Ventures) notes Ethereum's significant value and its impressive free cash flow production.

  • “Ethereum is worth 7 Snowflakes put together. And in q one, it produced 400,000,000 in free cash flow, which on a dollars basis made it the single largest producer of free cash flow of any publicly traded software company in the world.”

    Tomasz·0:00

  • “I love running this analysis, which is comparing Ethereum as a business to Mongo... Ethereum is worth 7 Snowflakes put together.”

    Tomasz·0:00
• $USDC - Mentions stablecoins' growing transaction volume and adoption by multinational corporations.

  • “Stablecoins now process 2 and a half times the volume of Visa on a monthly basis... There are large multinational corporations that are building their accounts payable and accounts receivable system on blockchain.”

    Tomasz·0:00
• $ALLIUM - Mentions Allium's thriving progress post Series A investment.

  • “We found Alium and Ethan. We referenced them quite a bit. I just done about 6 months researching, and then and then we preempted that series a, and the company has continued to thrive since then.”

    Tomasz·0:00
• $HELIUM - Mentions Helium as an interesting project.

  • “I think Helium is a very interesting project. The way that they build these networks and following that actually was running a miner for a long time in Helium.”

    Tomasz·0:00
• $DATA - Tomasz notes the potential of the web3 data ecosystem and how it could mirror the success seen in web2.

  • “Overall, we think that the data ecosystem is a great place to invest because there's a lot of compute used to analyze that data. And whenever there's a lot of compute, there's typically a lot of dollars.”

    Tomasz·0:00

  • “The second is that data volumes only ever increase. No one deletes data. And so Right. The amount to spend that will exist. The third is that the ability to create data products or insight on top or create alpha in the case of financial companies will always start with the data. So we think it's a necessary ingredient. But the point that you make around the contextualization is critical. Right? Just offering database of data is not useful enough. And the question is which workflows can be empowered or augmented or improved or automated with data. And there, I think it's still early. I think it's still early, and I think there's also it's still early in terms of understanding how do you combine Web 2 and Web 3 data in really intelligent ways to to achieve all those advantages that we just talked about. So it will come. And if you remember in the 2010s, there was this wave of big data. There was Cloudera and MapR and Hadoop. Venture capitalists and startups had this narrative that you should store as much data as possible. And because the cost of storage is low, eventually, you'll create value. And that's what happened for about 10 or 15 years until the dawn of the LLM era, where now we can finally make use of all that unstructured information. I think we're kind of in that phase in web 3 where we're gathering all this quantitative structured data. And some teams are able to use it for marketing analysis or alpha generation in the public markets, but the workflows are still not broadly yet understood because we haven't mechanized as an industry yet.”

    Tomasz·0:00
• $INFRA - Discussing web3 infrastructure as a viable investment

  • “I think the focus right now is on infrastructure, and it's a good place to be because the rate of progress in that infrastructure in terms of price performance is rapidly improving.”

    Tomasz·0:00

  • “The developer infrastructure, the tooling, and then analytical tools around it are great places to invest.”

    Tomasz·0:00

Bearish:

• $CRYPTOAI - Tomasz does not believe in the decentralized GPU model given security concerns and tech giant investments in the sector, which will likely respond to a GPU shortage.

  • “I am not a believer that Web 3 has a a major role to play there [in decentralized GPU clouds].”

    Tomasz·0:00

  • “I think there there have been lots of attempts or many attempts, and some of them successful of building decentralized GPU clouds primarily for training rather than inference. And that's, I think, a short term phenomenon that capitalizes on the supply chain challenges in the GPU stack. But ultimately, it's a difficult value proposition for two reasons. One is there are security holes within GPUs that are being fixed in the next generation of GPUs where if there are 2 tasks that are running on a GPU, one task can see the output of another. And so that's a big problem because if you have somebody running a very sensitive training job on a decentralized network, there's no trust. So that's a really big problem. I think the other challenge is everybody's investing in GPUs. Google, Microsoft, and Amazon are investing $12B to $15B a quarter in developing these data centers, not to mention Facebook, which I think will spend $20B on NVIDIA hardware alone this year. And so over time, the availability of the GPUs or the shortage in them will attenuate.”

    Tomasz·0:00