The internet is an inherently deflationary beast. Hydrocarbon and food (+ water) shocks are about the only effects that can overwhelm its economic power. The IMF may be a bit optimistic on those, but they are hard to predict. My only question in this is how strong the USD and Mexican peso stay.
The MSFT + OpenAI partnership is a big bet on semiconductor physics. There’s a significant chance that Azure won’t be able to host ChatGPT3 as a consumer-quality service. The Azure people are great; we use them at Pantastic. It is simply not clear that our species yet has the compute technologies to host a deep learning model that large in a way that its raw data is updated and can learn.
My expectation is that 100B+ parameter models exceed the capabilities of semiconductor physics in general, which means we are a ~decade away from being able to operate them.
Brands will probably waste their money this way – falling for the “first-mover advantage” pitch without understanding it yet again. However, consumers will buy products where the experience is most satisfying. Arbitrary digital addresses aren’t going to make the difference. Hyperlinks and their descendants are too easy to follow.
I’m not here to apologize for the horrific damage that Facebook has done to many countries’ political and social structures, including ours. Quite the opposite. Meta is a brilliant power grab by the oligarch formerly known as Facebook, and we all need to quit laughing at it. For reference, I got into the Facebook ecommerce ad business in the summer of 2007, weeks after it became possible, and have attached to it ever since.
Most observers miss two things:
- Meta’s existing dominance in digital goods advertising and sales (outside of China). Next to AAPL and GOOG, META is the only game in town and has a number of systemic advantages relative to those two phone OS makers. AMZN doesn’t hold a candle to META in this area, and NFLX is only dipping a toe in the gaming waters now.
- VR is BS and no one knows that as well as the owners of no-longer-called Oculus. Virtual Reality is just a marginal edge case of Augmented Reality (AR); it’s the intentionally private and public imagery laid onto our near-real life experiences that we will invest in. That Meta dystopia is designed from the start to be here in the “real world” – AFK and physically social – not in some virtual experience.
Meta has missed neither of these things; they are just running their normal playbook of distraction and confusion. They are moving to control what we see and buy in the overlay through which we many of us will be viewing our daily routines in ~5 years. Instagram Stores are a natural dominator of AR digital goods, viewed and bought through upcoming generations of camera-glasses.
The best-placed press (subscription) I’ve seen on Meta was in Business of Fashion. It is clearly just Meta’s PR department at work, but that’s exactly the audience that will invest in building and selling these digital goods as the rest of the fashion world goes pear-shaped. And those brands will get the buying and influencing started, all exactly as planned today.
A key feature of this metaverse concept is that the way we present ourselves, both in terms of our avatars and our clothing, won’t need to obey the norms and restrictions of [a purely] real world.
Idiot Test #6A: Founder Data Theft for LPs
This is an update to Idiot Test #6: Target VC Fund Size.
VCs broadly boast about both being data-driven and passing on 99%+ of the financings they consider. As I believe both of those data points, I do not regard VCs as investors. Also being data driven, I believe them to be Founder Data Thieves. Stealing founder data is what they do the huge majority of the hours in their work day, so I believe that is how they should identify.
In the specific case that they insist on being “stage-agnostic” or wanting to be “the first check” in financings that are far less than 2% of their current funds, the data theft is for the purpose of impressing their LPs. Later stage VCs (e.g. a16z) are in a race to tell their LPs what the investment sectors will be 2-3 years out before their peers are able to. They can only do that by seeing the private presentations of very early stage startups.
Please do what you can to avoid enabling this behavior. It only hurts the founders and early employees.
Quality early-stage investors will require you to scale a single revenue stream. If they ask about diversification, it’s to make sure that the answer is “no.” They are completely correct in this case. Diversification and fast scaling are incompatible. PG has a great metaphor for it:
A startup is like a mosquito. A bear can absorb a hit and a crab is armored against one, but a mosquito is designed for one thing: to score. No energy is wasted on defense. The defense of mosquitos, as a species, is that there are a lot of them, but this is little consolation to the individual mosquito.
Early stage capital raising is seasonal. To be most successful, start presenting either in September or (preferably) January, so that you have as long as possible before it is difficult for the partnerships to field a quorum for a full partner presentation. Getting a quorum generally becomes problematic in the US on July 1 and November 15. If you do not have a signed term sheet by those dates, the reasonable expectation is that you will need to re-start the process completely. And, you will be fundraising with a cloud hanging over your head. If you start late in the season (e.g. May 1), you set the upfront expectation that you are naive or desperate or both.
Here in early October, I am already deep into supporting several CEOs for the January 2019 presentation cycle. If you do not have a deck outline and rough milestones sketched out by now, you are late and risk being less effective than the other CEOs with whom you are competing with for investor attention. Please pardon the formatting on the schedule below; I copied it out of a Google sheet for the startup I came to London this week to support. US-based startups may be able to shave a week or two off the schedule by limiting their presentation geography a bit.
Raising capital at this level (<$20M or even higher) is also not about relationship building. It’s about showing that you know how to run a tight process that you will bring to the later rounds of financing – and FOMO. Stop having coffees and catch-up calls with VCs now. That is just giving them free information that you have sweat and bled for and ruining the showmanship of your presentation. What they care most about is that you are also the CEO to raise the several rounds of capital after this one without forcing them to do much work. The best ones may be very energetic and very helpful after investing, but they avoid making investments where they have no choice but to put in that effort.
I just wrote this boilerplate for the CEO here to in response to inbound emails. Feel free to use it:
Hi [xxxx],
Thank you for being so diligent in reaching out.
We are gearing up to raise our Series A in Q1. All of our first partner pitches will be the week of January 7 in London and NYC and January 14 in California. Between that and the quickly ramping sales of the Super Widget, I am not going to have time to chat.
We also are making sure that we target the firms we speak with pretty tightly.
Obviously [your fabulous investing entity] has the sector expertise we need, but may I ask the current fund size, age, percent invested to date, and reserve policy?
Thank you,
[the demigod of execution]
TGIM
October 22, Draft of deck with placeholders
October 29
November 5, Draft of deck with at least bad photos
November 12
November 19
November 26
December 3, VC target list finalized
December 10, Materials finalized Presentation deck, email deck, email text
December 17, Send out VC appointment emails
December 24
December 31
January 7, First GP meetings, London & NYC
January 14, First GP meetings, California
January 21, Second GP meetings, London & NYC
January 28, Second GP meetings, California
February 4, Full partnership meetings, London & NYC
February 11 Full partner meetings, California
February 18 Term sheets submitted
Pichai’s Predicament
My guess is that Google has a bigger problem than is obvious with regard to its under-development, PRC-censored Chinese search engine. The announcement of Google’s $550M investment in JD.com in June 2018 talked up “personalization,” which means Machine Learning. I would expect the contract to include an obligation for Google to provide search services in JD’s most important region – China. Search is a highly productive training component for ML which is a lot of what Google brings to the table for both JD and its conjoined twin Walmart. If I’m right, Pichai has already committed to shipping this service and has significant downside with both JD and Walmart against Amazon if he does not do so.
#conjoinedtwins !?!?
Review from old posts – JD’s board is dominated by Tencent and Walmart, the latter of which has a unified transpacific supply chain with JD. Also, Baidu is more closely affiliated with Tencent’s main Chinese rival Alibaba, so there is a real role for Google China in this alliance.
Walmart Continues their Sad Trend
As with the Jet and Flipkart acquisitions, Walmart keeps buying market share and its associated consumer data but not the machine learning talent needed to turn it into a sustainable advantage. Building a world-beating tech team requires more than great online retail talent (all three acquisitions have that!), and more than throwing bodies at the problem. It requires the kind of leadership that Walmart can’t organically assemble.
Worse yet, Walmart’s two dominating machine learning partners, Tencent/JD and Google, are now ganging up. I can not guess exactly how Walmart is going to be endrun by those two, but I do not think it will do much for Walmart shareholders. Google is already strong in Latin America, and Tencent has quietly started increasing their investments there. I have been doing increasing work in the region and the Tencent presence gets thicker by the month.
Statistically, VCs are not investors. They are analysts who steal founders’ information 99%+ of the time in order to use it elsewhere. As with most of the technology startup world, they “suck less” so they are by far the best we can produce.
That all means that their workflow is optimized to pass, which is not in your interest. “Respect their time” and similar comments are exactly the wrong thing to obey if you want their investment. Be the exception.
I love Mark, but he’s speaking only in his interest, not any founders’. Your deck is not your best marketing tool, you are. And, add tons of purchase friction. Otherwise, you are easy come, an easy free education, and easy go. You don’t stick out, maybe you said it in person for 45 minutes, maybe you didn’t get the meeting at all, and – pass.
My apologies to the anonymous donor who sent me this email but:
I know the deck is lame, and I know I’m not supposed to throw it “over the transom,” but it drives me crazy to see so many lame business models get funding when we have a real business that solves a real business problem and we can’t click with any VCs.
Don’t let it drive you crazy, just recognize that it’s secondary at best and irrelevant at worst.
Instead, it is great fundraisers that get funded. It’s occasionally much easier if they have a great business, but it’s not the primary criteria. It is also nothing to get angry with the VC community over. No matter a VC’s personal goals, their very specific job is to buy private stock at a low price and get cash or a liquid public security for that same stock at a much, much higher price. If they can chase their pro rata several times in the interim, then they might make a good amount of money on the investment.
In addition, the incredible volume of deal flow volume processed by professional venture capitalists nearly guarantees that they can not possibly understand the businesses they invest in until the fifth or sixth board meeting. That is especially if a fundraising was at all competitive for them to participate in.
That means their first screen mental is, “Without a ton of coaching from me, can this CEO survive my fund’s partner meeting and, in a year-ish, comfortably raise money from the next set of investors? If so, she will raise stock price, open up a pro rata investment opportunity for our fund and maybe pay for my kids’ tuition in a few years.”
Great fundraising has to drive the design of your deck, not some misguided attempt to teach people about your business. Make the audience believe in YOU as the leader of this enterprise; nothing else works.
Every major technology wave changes distribution in some significant way. Machine learning (more popularly “AI”) is no different; having the largest relevant training set wins. It will prove powerful enough to drive vertical integration for the data set owners/managers. That means that “brand” is less important than data collection during the ~5-year land grab we are are now working through. Direct, branded paths to the consumer will re-emerge as the highest value, but that will just be the winners cleaning up maturing, consolidating markets.
I arrived at this understanding anecdotally. I keep my focus entirely on ecommerce infrastructure and logistics, so I am only involved in three ML-at-scale startups. The founders of the two most mature both clearly have reasons to see B2B2C white-labeling as a reasonable and high value option to gather consumer data. Given my last 20+ years of internet experience, that was confusing at best and I started off biased against it for all the reasons that venture investors will be in terms of realizing shareholder value. Well, I was wrong; the head of the power law will be awarded to the systems that learn the most the fastest – no matter where in their lifecycle the end-buyer becomes loyal to them.
As a corollary, any time distribution changes, some VC pattern-matching is disrupted. That’s true here as well; and the VCs that figure out the new pattern sooner will make more money. I should do an Idiot Test on VC pattern-matching sometime soon.