Subscribe
Share
Share
Embed
A debut catch-up on what your newsletters have been saying: the power grid as the real AI bottleneck, whether the boom is 1999 again, and two truths about how businesses actually grow.
My weekly email knowledge dump
This is MailDump for the week of August third, the first proper one. I went back through the whole backlog in your inbox, so treat this as a catch-up on what your newsletters have actually been saying. Four things kept surfacing: a quiet obsession with the electrical grid, a belief that clean energy is winning even as the politics get uglier, an open fight over whether this AI boom is nineteen ninety-nine again, and, twelve years apart, the same two truths about how businesses really work.
Start with the grid, the single cleanest agreement across everything you read, from people with nothing in common. The claim is blunt: what is throttling the AI build-out is not chips, it is electricity, and specifically the wires. David Roberts at Volts points out the US grid runs at only about fifty percent of capacity on an average day, because it is built for its single worst hour; his fix is to squeeze more out of what already exists, through dynamic line ratings, better conductors, and demand flexibility, before stringing new lines. The newer problem is the data centers themselves: an AI data center can drop hundreds of megawatts in milliseconds, a demand the grid was never engineered for, which the operators call inertia-less load. The rushed response is to bolt a private gas plant onto each one. Former energy official Jigar Shah calls that the dumb way, and argues for flexible interconnection instead, letting the data center plug in fast in exchange for being curtailed during stress, backed by battery storage. And the United Nations puts the global number on it: at least three thousand gigawatts of renewable projects sitting finished or planned in connection queues, stranded not by any shortage of panels but by a shortage of transmission.
Here is where they stop agreeing. They share the diagnosis, wires, but split on the cure. Roberts and Shah are effectively arguing against building: wring more out of the grid we have and sidestep the five-to-ten-year queue. The United Nations is arguing for building: three thousand gigawatts are stranded because the transmission does not exist, so spend, roughly a trillion dollars on new grid this decade. The detail that settles it for now is time, because new transmission takes the very five to ten years that created the jam, which is the strongest case that the near-term win is squeezing the wires we already have, even if the UN is right about where the decade has to end up.
That grid is already carrying more clean power than most people realise, which is the closest your inbox comes to good news. Katherine Hayhoe's sharpest data point: solar has overtaken coal on the US grid for the first time on record, twelve point eight percent against coal's twelve point two, though she calls it an inflection, not a victory, since coal outweighed solar five to one only five years ago. Globally it is starker, an Ember analysis found renewables outgenerated coal worldwide in twenty twenty-five for the first time since nineteen nineteen, by crowding coal out rather than because demand fell. China sits underneath that, emissions flat or falling for twenty-one straight months while solar grew forty-three percent year on year. And storage is following, one hundred and twelve gigawatts of grid batteries installed last year, more than triple the year before.
The real disagreement here is not direction, it is autopilot. The milestone-counters read each record as momentum that feeds itself. Hayhoe and Roberts push back from the same spot: the economics have already won, cheap solar and storage now beat coal on price, but the politics and the pace have not, and Hayhoe's own number, clean-energy investment growth slowing from twenty-seven percent in twenty twenty-one to eight percent now, is the evidence it does not run itself. Roberts puts the cleanest line on it, that the clean-energy story and the climate story have come apart, the first winning while the second gets harder. The honest read is that clean energy wins on cost and loses on speed, and closing that gap is a political job because the brake now is not the technology, it is permitting, transmission, and the same interconnection queue from a minute ago, and every one of those is set by legislators and regulators rather than engineers. You can see it in the split Hayhoe keeps flagging: where the politics leans in, like the states still expanding even as Washington cancels funding and rolls back the rules, the build keeps pace; where it leans out, the cheap clean power just sits in the queue.
All that build-out is where your inbox stops agreeing and starts arguing, over whether the whole thing is a bubble. Scott Galloway went fully bearish in July in a piece called nineteen ninety-nine dot A-I, tracing how the dot-com collapse moved in dominoes, from consumer flops like Pets dot com, to suppliers like Sun Microsystems, down ninety-six percent, to infrastructure firms like Nortel, over ninety percent erased, once their vendor-financed customers went bankrupt. In a separate edition he needled the hype from another angle, noting a humanoid robot that earned its place in a Beijing half-marathon only after falling and being carried mid-race, his evidence that the spending is sprinting well ahead of the capability. Benedict Evans threads a stranger needle he calls the deflationary bubble: it can be genuine overinvestment and speculative froth while also driving down the cost of thinking itself, so both are true and only the sequencing is open. And Dror Poleg gives the vertigo version, an upside-down pyramid where growth rests on data-center construction, which rests on five companies, which rest on two startups, which rest on a handful of researchers who, in his words, do not fully understand their own models.
For all the heat, none of them actually disagree that the money is excessive. Where they split is on what a reckoning would destroy. Galloway expects a two-thousand-one-style wipeout that takes the companies down with their valuations; Evans expects almost the reverse, a crash that clears the froth but leaves the cheaper thinking it paid for still standing; Poleg says it barely matters which, because the real exposure is the narrowness of the pyramid. So the frame that actually helps is not bull versus bear, which all three reject, but the single question their argument really turns on: whether the unwind, if it comes, erases the capacity or merely the price of it.
Now the pattern I find genuinely lovely, because it repeats across twelve years and two different eras. In twenty fourteen, in a Stanford lecture, Facebook's head of growth Alex Schultz drew a hard line between growth that looks real, acquisition, and growth that is real, retention, calling the alternative a leaky bucket that no acquisition budget can ever fill, and telling founders to graph their day-one, day-seven, and day-thirty retention before spending a dollar. Eleven years later Elena Verna, writing from inside the AI company Lovable, says the identical thing in an essay titled Revenue Addiction Kills Companies, adding that it is now fatal rather than merely wasteful, because switching costs are collapsing and a better rival can appear weekly. Alex Hormozi puts a threshold on it: keep spending on marketing until word-of-mouth outruns churn, and only then have you hit escape velocity.
What is worth noticing is that they hold the same principle but have quietly swapped the prescription. Schultz's twenty-fourteen answer is a measurement discipline, graph the curve and let it tell you whether you have earned the right to spend. Verna's twenty-twenty-six answer is an emotional one, make people love the thing, because the features are now commodity and a good-enough rival is one click away, so attachment is the only moat left. Hormozi sits between them. The meta-lesson is that the rule survived twelve years intact while the lever underneath it moved, from measuring retention to manufacturing it, and what moved it was the collapse of switching costs.
The other business truth your inbox reached twice is about what a company is actually worth, and it produced the single sharpest number in the pile. Timothy Armoo, who sold his own agency in his twenties, ran a report called Seller Secrets, twenty-seven founders and just over three hundred and sixty million pounds of real exits between them. His standout case is a medical-supplies business that sold for twenty-eight million pounds, while direct competitors on the same revenue and margins sold for around seven. The entire gap came down to whether the founder was optional. His diagnosis is precise: when thirty to forty percent of revenue is tied to relationships that only the founder holds, that is the biggest red flag a buyer sees, and it can roughly halve the valuation. Andrew Wilkinson, who buys and holds businesses at his company Tiny, arrives at the same place from the opposite chair, rejecting any business that needs what he calls operational heroics, because, in his phrase, heroics don't compound, and hiring people who can run the thing without him.
So here is the actual list the two of them describe, and it is the same list whether you are selling or buying. Document the processes, so the business runs on a system and not on the founder's memory. Move the client relationships onto the team, deliberately, over twelve to eighteen months, so no more than a sliver of revenue depends on a person who might walk. Turn one-off projects into retainers, so the revenue reads as recurring. Systematise the delivery so a new hire can run it without the founder in the room. Build a predictable pipeline instead of a founder-driven one. Structure any eventual deal to avoid an earnout that quietly re-chains the founder to the desk. And treat diligence as the real exam, because that is where a founder-dependent business gets found out. That is Armoo's seller's checklist, and it is, item for item, what Wilkinson screens for as a buyer and rejects when it is missing. Read it forwards, it is exit prep; read it backwards, it is an acquisition filter; either way it is just a description of a business that does not need you in the room.
One last thing to leave you with, from the health and performance shelf, because it cuts against the usual more-is-better noise. Across dozens of interviews on the Finding Mastery show, the recurring surprise is how small the effective dose actually is. The attention researcher Amishi Jha puts the floor for training your focus at twelve minutes a day. For your heart and lungs, the protocol that keeps coming up is two twenty-second, all-out efforts, twice a week. The addiction psychiatrist Anna Lembke resets a burnt-out dopamine system with a single thirty-day fast from your compulsion of choice. These are different experts in different fields, not people in dialogue, so there is no argument to settle; what is striking is that they independently converge on the same shape, that the effective dose is tiny. The one thing pulling against them is that each is a single voice on thin evidence. So the honest meta is to treat the small dose not as a proven protocol but as a low, inviting floor: the cost of testing whether twelve minutes, or forty seconds a week, does anything for you is almost nothing, which is rather the point.
That's your inbox, caught up. See you next dump.