Dan Ramsden

Markets, other networks.

[Dan Ramsden @d_ramsden]

@AlistairCarns May not be a matter of money… unless you meant “cost” more comprehensively.

[Dan Ramsden @d_ramsden]

@priceactionlab It's not unlikely he'll be a billionaire again for a while, so maybe things will settle down.

[Dan Ramsden @d_ramsden]

@IanBegley What would Pat Riley do?

[Dan Ramsden @d_ramsden]

@dampedspring 100%, and it makes one wonder why analysts are so credulous most of the time, or media consumers. Maybe that too is a lie.

[Dan Ramsden @d_ramsden]

@DannyDayan5 Yup, stocks down while rates, oil, dollar all down also. Maybe funds are being raised for the coming IPO?

[Dan Ramsden @d_ramsden]

@priceactionlab Yup, we better hope...

[Dan Ramsden @d_ramsden]

"... instead, he argues they behave more like complex adaptive systems, shaped by feedback loops, human behavior, volatility clustering, regime change and recurring fractal patterns."

Like social networks, but more...

[Keith McCullough @KeithMcCullough]

Want to learn about Fractals in markets?

[Dan Ramsden @d_ramsden]

Yes. And the worse news is that valuing assets is close to impossible in a rapidly evolving and transformational environment. Exceptions are businesses with high optionality and/or (mostly and) high likelihood of a continuing and stable core. This may explain the relative attraction of "hyperscalers" the past few years.

[Dan Ramsden @d_ramsden]

And on top of it all, how is the terminal value justified when the useful prediction must not only capture long-term growth and profitability potential but, more importantly, the very nature and profile of the business (which is rapidly changing) into perpetuity?

[Ptuomov @ptuomov]

HOW TO ABANDON THE PRESENT VALUE FORMULA IN POLITE COMPANY

While retail investors are free to completely abandon the present value formula whenever they so desire, professional sell-side research analysts do not have such freedom. The professional standards force those analysts to follow a set of complex rules and social conventions resembling a tea ceremony to do so.

The research analyst has a target price that has to be about 15% above the current price. Then he must construct a set of cash flow forecasts, long-term growth forecasts, and discount rates that mechanically justify that target price.

In a bubble, the price is unjustifiable, which means that those forecasts must also be unjustifiable, but they must appear on the surface to be as justifiable as possible. Furthermore, the near-term cash flow forecasts must actually be relatively accurate because the reality of those near-term cash flows will, by definition, be revealed in the near term.

The two main ways analysts can tune their present value formula to justify the unjustifiable target prices are (1) pushing out the earnings in the multiple and (2) increasing the long-term growth rate. The first method simply says that the front page of the report will not compute multiples based on year 2026 or 2027 earnings, but year 2030 or even year 2040 earnings. This is relatively safe, as both the analyst and the institutional investor listening to the analyst are likely pursuing other career paths by 2040, when those 2040 earnings fall short of the forecasts.

The second method is just to increase the terminal value and terminal multiple after the explicit forecast horizon by increasing the long-term growth rate forecast. Owen Lamont has recently written about this, but the observation that the analyst long-term forecasts become unrealistic in a bubble is almost as old as the field of security analysis itself as each analyst covering each stock has to stretch the long-term growth forecast higher and higher.

Although some stocks may meet these high long-term growth forecasts, at some point of the bubble they will aggregate for the whole market to a level that is almost certainly impossible for even the godliest Machine God to produce. This observation is also not original but has been made, among other people, by Cliff Asness in his “Bubble Logic” piece.

A good proxy for Step 4 of the bubble is to compute the difference between aggregated individual stock analyst long-term growth forecasts and macroeconomic analysts’ long-term GDP growth forecasts. (This may be the sole case in which macroeconomic analysts’ forecasts of anything have any utility.) When the bottom-up LTG aggregated across stocks is unusually high compared to the long-term GDP growth forecast, that is evidence of the professional investors taking the fourth step and abandoning the discipline of the present value formula in a polite way.

[Dan Ramsden @d_ramsden]

@BobEUnlimited What data is there to support that the trend you show would have continued (as you show) if not for AI?

[Dan Ramsden @d_ramsden]

@DannyDayan5 @NickTimiraos The myth of "data dependence", which is in fact "interpretation dependence", which is itself a distortion of so-called "independence"... as the dependence is on one's particular bias, which can be influenced by so many things (including politics).

[Dan Ramsden @d_ramsden]

A thesis, based on history and the latest:

1) The tokenmaxing/pricing issue is part of the market mechanism for new offerings whereby the price eventually settles at a clearing rate based on supply/demand factors. In highly liquid and mature markets this happens quickly, in new and less liquid markets it takes a while. In this case, it will take a while, but it will happen. One variable that complicates matters is the fast and continuous improvement of the product (on the supply side) vs. a still unknown and possibly changing need for such improvements (the demand side), which will take additional time to resolve.

2) The commoditization of the product doesn't negate a winner-takes-most mechanism, it only suggests that the winner(s), especially in light of pricing uncertainty, will as always be at the network/distribution layer rather than the product layer... in other words, the ones with greatest flexibility, market reach, and long-term survivability, including the ability to subsidize offerings with other revenue streams from a captive audience. Most of the so-called hyperscalers are in this category. The pure-play operators about to IPO, less so, but may join the fray with adjacent offerings to emulate, say, Alphabet or Amazon... or Verizon... or SpaceX, for instance. None of which will be easy. Very difficult, in fact. Close to impossible.

[Dan Ramsden @d_ramsden]

@DannyDayan5 @DeItaone Was misquoted, the actual quote is "through highly placed representatives"

[Dan Ramsden @d_ramsden]

@EPBResearch Not conglomerates… multi-tiered, multi-directional network platforms… This is a critical distinction because, unlike for the old conglomerates, in the current case the underlying interconnected systems make the parts as well as the whole extremely difficult to disrupt.

[Dan Ramsden @d_ramsden]

@dampedspring @citrini I think a bubble can also be seen as a wide divergence between technicals (supply/demand) and fundamentals (inherent asset value)… If so, we may have a bubble for some but not uniformly. In many important cases price seems justified by fundamentals.

[Dan Ramsden @d_ramsden]

Great interview and strong thesis. Something I wonder about: If you strip out tech CAPEX the economy may actually be recessionary now (stagflation). A rate hike may not be sufficient to slow investment down at this stage, and bring down inflation, but only make matters worse for the rest of the economy, including an impulse towards AI driven layoffs to reduce costs.

[Dan Ramsden @d_ramsden]

@DannyDayan5 You may have skipped a step, go back to universities, where both the readers and the writers are trained. In fact, maybe even further.

[Dan Ramsden @d_ramsden]

Outstanding analysis.

[OM @om]

How AI Is Changing the Network(s)

As is always the case, this started with a simple question: Will AI change how networks work? Will it impact the speeds we need at home and on our phones?

My assumption was that AI would accelerate this — personal AI agents querying the cloud all day, your house talking constantly to a model (or models). A lot of this is still wishful thinking.

My attempt to find an answer led me down a whole new path of inquiry, with surprising results. The real action is happening far away from the madding consumer crowds. None of this was surprising, considering I have covered the evolution of the internet and its innards since the early 1990s.

Internet 1.0, Internet 2.0, the cloud, mobile, data and machine learning, and now AI are all part of a continuum that has challenged and scaled the network, helped evolve new technologies, and introduced new ways of thinking about ever-expanding oceans of data.

AI is only supersizing everything, including the sheer scale of capital it needs to build competitive advantage. It is also bringing down the curtain on some of the old ways of thinking about the cloud, data centers, and networks. If my old publication were still around, we would be writing about all of this and more.

Over the past few weeks, with help from old friends in the networking and infrastructure world, I have managed to put together an almost 5,000-word overview of the changes to the network. I look at the physical pipes, the shifting demand profile, and who really owns this new internet.

It is by no means complete — nor is it meant to be. Instead, it is an anchoring essay for you to think beyond the dollars and the hype machine of AI, and see that this is just tech doing tech things at a scale we have never seen before.

Continue reading my essay, "Say Hello to the Internet of AI."

https://t.co/IVcFWFp6Ag

[Dan Ramsden @d_ramsden]

@DavidSacks @pmarca “At this point, stopping progress in AI would be equivalent to halting the U.S. economy.”
Isn’t this actually a macro risk factor?