Les Prophéties du Porch(dog)

The greatest fortune tellers have several things in common: (1) they are extremely vague so that they are always right somehow and (2) they are still wrong a lot. Consequently, I consider myself among the greatest fortune tellers of technology. With that in mind, let’s have a look back at what I wrote over the last dozen years and what I got right and what I got wrong….

But first let’s start with the (first) big call that I got right a long time ago (in a galaxy far far away). And why not start off on a good note….. In 1997 or 1998, when I was in investment banking, I wrote a report that showed that ~100% of American homes had computers, but only ~20% of them were connected to the Internet and ~100% of new computers sold were sold with Internet connections so that gap would close to zero. And it did. Yada yada yada… Internet bubble.

Theme #1: The Facebook model (but drop the “The…”)

It would be more than a decade until I would start writing again. (January 7, 2011, to be specific.) I had been hopped up on the “this is happening again” concept of (what we would come to call) Web 2.0 and wanted to put a stake in the ground on the off chance I was right. So I penned this (first real) post about Google and Facebook in March 2011. The primary premise was that Google should and would be threatened by the emergence of Facebook et al. Not that Google would go away (any more than Microsoft did when Google emerged), but that ad dollars would disproportionately grow toward social networking from (pure) search.

Verdict: Correct.

In fact, to put my money where my mouth was, I started buying FB shares in October 2012 at ~$20/share and continued buying over the next year mostly between $20-30/share and also in the $40s. Current FB stock price: >$350/share. Brilliant, right?

Here’s the rub(s): (1) I started selling the stock a year later between ~$60 and ~$100/share ; (2) of course the growth rate of ad dollars to social networking would exceed comparable figures for (legacy) search - it was from a base of basically zero so that call was kinda baloney; and (3) my contention was that money would flow more profoundly to social networking because companies like Facebook and Twitter would do search better than Google given the “social lens” that would make results dynamic and more relevant. I doubled down on this search thesis in April 2013 and plenty more times (than I care to admit) on and offline.

Verdict: dead wrong.

Facebook et al never really bothered with search (and it doesn’t seem like they ever will); rather, the social networking companies more effectively surfaced “relevant” content for users, which made them more likely to engage, which made the content more likely to generate traffic and therefore ad dollars.

Where I was right, however, was about Facebook et al’s business models. The technosphere and Wall Street and TV talking heads enthusiastically scoffed at Facebook’s seemingly absent business model and its seemingly ludicrous valuation (which turned out to be retrospectively ludicrous in the other direction). And they smugly declared that the move to mobile computing was the end of Facebook. In December 2011 and again in July 2012 (and plenty of other times that I’m forgetting but insist actually happened because it makes me sound smart(er)), I wrote about the evolving lucrative promoted tweets / sponsored stories / etc. business models that would be catalyzed by the move to mobile and would ultimately catapult Facebook to a nearly trillion dollar valuation (and Twitter into the loving arms of its new abusive overlord).

Verdict: nailed it.

I also predicted in June 2014 that lots of revenue would come from messaging (it does) and that Facebook’s cuckoo acquisition of Oculus might seem prescient at some point (jury’s still out, but that Apple Vision Pro thing looks pretty rad).

Verdict: ??.

Theme #2: web / tethered / local → social / mobile / cloud.

The really big call I made, which was comparable to the circa 1997 Internet (Web 1.0) call, was in a series of posts in July 2011 about the move to social, mobile, and cloud computing (which we now call Web 2.0) and its impact on markets and economics. The series (dubbed “It’s the (Social/Mobile/Cloud) Economy, Stupid”) was a three-parter starting with (1) overall thesis (i.e., social / mobile / cloud), (2) the economic impact, and (3) the stock market impact. And with quasi Nostradamian flair, the overall thesis was (sort of) a quatrain (in that it had four parts, though none particularly poetic) - copied / pasted here:

(1) Web → social computing, tethered → mobile computing, and local → cloud computing equals big times three.

Verdict: Four-Peat. (Trademarking that; suck it, Pat Riley.)

• There are almost 5 billion people using social networking now (from 1B+ in 2011 and, like, zero a few years earlier).
• Net jobs numbers from 2011-2016 below (from BLS / ChatGPT, which I didn’t proofread).2011: 1.561 million2012: 2.467 million2013: 2.203 million2014: 2.560 million2015: 3.104 million2016: 2.506 million
• NASDAQ levels at the end of 2011-2016 below. (Fwiw as I write this it’s >14K).2011: 2,605.2012: 2,960.2013: 4,154.2014: 4,777.2015: 5,066.2016: 5,383.
• Consumer spending grew at a ~3.5% CAGR between 2011 and 2016 for total growth of almost 20% over that time period. Easy one (because the general trend is almost always up and to the right), but still:2011: $10.7 billion.2012: $11.0 billion.2013: $11.4 billion.2014: $11.8 billion.2015: $12.3 billion.2016: $12.7 billion.

Moreover, I stayed on top of this call and doubled and tripled down in February 2012 and May 2013 when the “consensus” view was that valuations were insane and the bubble would burst. In 2012 my rationale was that there was lots of cash on the sidelines and nowhere to put it other than private growth companies; and in 2013 my rationale was that the big moves were still coming (smartphone usage would double from 50% penetration, social media would penetration would go from 30% to ubiquity, mobile traffic was eclipsing desktop, cloud was teeny, etc.) So I’m taking credit for this call - it was a huge one and it’s the huge ones that count :)

Theme #3: “Other.”

This is where it gets a little dicey ??

• I was and remain convinced that we’ll ultimately use mobile devices to find, book, check-in, order, pay, and return to restaurants and other retail establishments, but my calls on Foursquare being the winner in that space (in November and December 2011) were woefully wrong. I suppose I can claim some redemption naming Uber as a quintessential mobile payments structure in January 2012, but I’m grasping for straws.
• In August 2011, I wrote a counter argument to Groupon naysayers and said that despite not being a buyer I was warming up to it, but I’m using my Get Out of Jail Free disclaimer and hedge to get out of that one.
• I made a (bullish) YouTube call in April 2013 that I think was mostly right, though honestly it was well on the way by that point.
• And I still want to see a reality TV show called “Ready? Set? Code…” that I talked about in September 2013, but I don’t think that’s showing up in the EPG any time soon. That said, I believe the thesis that “Coding is 21st century literacy” continues to come to fruition.

Verdict: mostly wrong, but some silver linings.

I didn’t write much for the next ten years other than how to work with me on intros and how to use Amazon’s leadership principles to find a nanny as well as post mortems on founding Tigerbow and Philo (the latter of which I used as my Amazon interview essay) because I got jobs to co-build the startup teams at AWS and Stripe. But with those missions “accomplished” I picked up the pen again. So let’s see how we’ve done lately.

Since early last year I’ve been writing about a bunch of themes in and around Web3, blockchain, AI, metaverse, GPUs, and the like. To me, they’re all related and fall into one big(ger) theme (much like social / mobile / cloud did), but let’s dive into the subcomponents one at a time; those being: (1) blockchain, (2) GPUs, and (3) AI etc.

Theme #4: Web3 / blockchain.

In 2011 I wrote something like “It ‘feels’ like this is all happening [again]” so in March 2023 I started off with my grand thesis titled “Ah Sh*t, Here We Go Again….” because it certainly feels like we’re in the midst of another gigantic platform shift. (Not sure anybody got the double entendre on the title - it’s a quote from Grand Theft Auto, which, for the purposes of these posts, I’ll consider a metaverse ;-)) Did I get it right? Kinda yes, kinda no. My thesis has been that the next move is to blockchain / AI / metaverse and is being catalyzed by GPUs, but that first post focused more on the blockchain component. I still have (very) strong conviction that blockchain is among the fundamental technology shifts catalyzing this move, but the application stack rewrite is more for AI than blockchain at the moment.

So was I wrong? Kinda yes, kinda no.

First, I followed up in April 2023 with a piece explaining my definition of Web3, which conveniently included AI (and metaverse) more prominently. I still believe that, but it’s clear that the general consensus associates blockchain with Web3 and not the other components. Touché. But I did conjecture that these three seemingly unrelated trends had one common thread: massive compute requirements and that Nvidia was both the catalyst and the beneficiary of this phenomenon. Moreover, to put money where my mouth was (and remains), I bought NVDA shares between December 1, 2021, and March 9, 2022, at ~$225 to $330/share. It was a bumpy ride, but the stock currently sits near $500/share making it about a double and my largest (personal account) holding. Brilliant, right? I also bought RBLX (down ~30%), COIN (down ~10%), and a couple of others that are way (way) down and nearly offset my NVDA gain (but didn’t - phew).

Second, I still believe that blockchain will have its place in both AI (authenticity) and metaverse, which I wrote about last August, but I did concede that “[c]rypto is currently useless (for consumers) in the US aside from currency speculation… and buying drugs….” That said, crypto has indeed found (massive and growing) utility in financial markets that you just don’t see unless you’re deep into the overnight lending, commercial paper, repo, and other markets. For example, JP Morgan’s Onyx platform is transacting up to $2B in volume per day. Why? Because “[p]utting assets on chain could revolutioni[z]e traditional markets by allowing instant settlement and the ability to rapidly transfer assets, reducing risk and increasing liquidity, especially for illiquid assets such as private markets.” TLDR: it’s cheaper and lower risk to setup, execute, and settle these transactions than using traditional financial networks. Note: the commercial paper market is ~$1T and the repo market is $2-4T; these are some of the markets that lubricate the everyday functions of businesses (like payroll, inventory, etc.)

Verdict: mixed / too early to call. ??

Theme #5: GPUs are eating the world.

It’s the move to GPU computing that has been underpinning this whole phenomenon (in my opinion). I speculated last year with no data (and still have no data) to support that we hit a tipping point in price / performance for compute that made AI, blockchain, etc. not just technologically feasible but also economically feasible. Thank you, Nvidia. But I had been looking for (i.e., cherrypicking) a(n over)simplified set of numbers that would support my thesis - similar to the PC / Internet gap of the late 1990s or the doubling of the <50% smartphone penetration and increasing allocation of IT spend to cloud (above 10%) that would happen in 2010+. I finally “found” that number in June of last year and wrote about the attach rate of GPUs to CPUs growing from 10% to… lord knows what, but way (way) higher. (See my deliberate ambiguity? That’s how the crystal ball is always right.:)) TLDR: right now, there 400 million CPUs and 40 million GPUs sold per year and every GPU requires a CPU so that attach rate is 10% (ish). And because GPUs are orders of magnitude more effective at training AI models etc., the growth in GPUs (and therefore GPU attach rate) will be… large.

Verdict: so far, I’m right. ??

(And finally) Theme #6: AI business models and other stuff.

In the spirit of my posts a decade ago re people not understanding Facebook’s and Twitter’s business models (and previously not understanding Google’s business model ten years before then), I wrote in August 2023 about business models for AI etc. more closely resembling Spotify than Google etc.; i.e., rather than sponsored content (in search results or feeds), OpenAI and others would “pay” providers for using their content to train and run AI models and that money would come from revenue (subscriptions and ads and other) from users “paying” for access to better models; similar to how Spotify has a free tier and paid tiers with more / ad-free music, and the artists are compensated from that revenue depending upon how often their content is consumed.

Ok so where are we on this? Sarah Silverman was the first creator I recall suing OpenAI (and others) for training models using her content though I’m sure there were others. More recently, the NYTimes sued OpenAI and Microsoft for the same sorta stuff and I’m sure there are others here as well. Pretty soon you won’t be able to swing a dead cat in the media industry without hitting someone suing OpenAI - that wasn’t and isn’t a surprise. But the first step I saw supporting my thesis that these companies will work together was Axel Springer partnering with OpenAI last month to improve AI usage in journalism. (Axel Springer owns tons of brands like Politico and Business Insider.) Is this the exact arrangement I predicted resembling how Spotify works etc.? No. But we’re moving in that direction.

And because of my expectation of the forthcoming romance between OpenAI et al and content creators, I speculated last September that traditional SEO will evolve to help creators ensure their content is used to train and run AI models. It’s not a thing yet, but people are talking about it aside from me. And even ChatGPT told me that “…the concept of deliberately optimizing content for inclusion in AI model training is an emerging area and not as well-established… as SEO…. However, there are several reasons why… groups might be interested in this approach….” (And I think it’s pretty clear what those reasons are.)

Verdict: still early, but….

So there you have it. In summary….

• 2011-2013:I was right about the (giant) move to social / mobile / cloud and its impact on tech and economics and financial markets. And I was right about the power of the business models for social media companies.But I was wrong about Facebook search, Foursquare, and a bunch of other individual companies.
• 2014-2023:AWS got (really) big, and so did Stripe.So I’m giving myself credit for being in the right place at the right time.
• 2023-present:Got off to a bumpy start re the move to blockchain / mobile / metaverse with too much focus on blockchain / crypto, but….So far so good on GPUs and AI and AI models; and blockchain / crypto is picking up steam in (really) big, (really) boring markets where it is having a transformative effect.

And now, back to the future….

(For more like thism, subscribe here.)