Author: philg

Years until NVIDIA’s revenue exceeds U.S. GDP

by philg

Happy Labor Day to those not smart enough to refrain from working!

Soon all of our dreary poetry writing, creative image and video production, etc. will be accomplished by LLMs. Freed from this labor we will be able to spend all of our time cleaning the kitchen, unclogging drains, fixing roofs in 90-degree Florida heat/humidity, etc.

Let’s check in with the machines that will be doing all of our laboring going forward.

NVIDIA recently released some financials:

My question for NVIDIA chips, via ChatGPT:

Extrapolating from revenue trends of the last five years, how long will it be before Nvidia’s annual revenue exceeds the United States GDP?

The answer:

Short answer: about 2036—but only under a wildly unrealistic straight-line extrapolation.

Here’s the quick math I used:

  • Nvidia’s last five fiscal-year revenues (billions): 2021 $16.7, 2022 $26.9, 2023 $27.0, 2024 $60.9, 2025 $130.5. That implies a ~67% CAGR from FY2021→FY2025. Macrotrends
  • U.S. nominal GDP: about $29.2T for calendar year 2024 (BEA Table 1.1.5); latest quarterly SAAR (Q2 2025) is $30.35T. FRED+1

Assuming Nvidia keeps compounding revenue at ~65–70% annually and U.S. GDP grows ~4% nominal per year, Nvidia’s annual revenue would mathematically catch U.S. GDP in roughly 11.4 years—i.e., mid-2036. With GDP held flat at today’s level, the crossing is ~10.5 years. (This requires ~224× growth from FY2025 revenue.)

Sensitivity (still using simple exponential curves):

  • If Nvidia grows 40%/yr → cross in ~18 years (≈2043).
  • 30%/yr → ~24 years (≈2049).
  • 20%/yr → ~38 years (≈2063).

⚠️ Reality check: this comparison is apples-to-oranges (company revenue vs a country’s value-added), and no firm can sustain ~60–70% annual growth for a decade+ in a competitive, capacity-constrained industry. Treat the dates above as a fun extrapolation, not a forecast.

How did Grok do on this question? Terrible. Its initial interpretation of the question was that it was about this year’s U.S. GDP, i.e., that it should hold GDP constant. When corrected, it chose an absurd 8.5 annual growth rate for U.S. GDP (would make sense only if there is 6 percent inflation; we’re lucky if we can eek out 2 percent real GDP growth even after flooding the U.S. with migrants; per-capita real annual GDP growth in the U.S. has been about 1.3 percent per year during the past 25 years (FRED)).

Loosely related:

Full post, including comments

Could we shade the Earth or at least some cities now that Starship is working?

by philg

Elon Musk wants to go to Mars on Starship. What can the technology do for those of us who aren’t interested in living on Mars? How about improving our beloved Earth, or at least the biggest and richest cities, via space-based sunshades? If we wanted to help everyone on the planet we could put a 35kTon shield at the L1 Lagrange point (#Science).

But if we’ve learned nothing else from the Climate Change (TM) experts we’ve learned that the climate righteous don’t care about everyone on the planet. Greta Thunberg, for example, cares only about the noble Palestinians (today she sails to Gaza from Spain). Climate change alarmists in New York City were in favor of open borders when migrants invaded Texas and against open borders (“a crisis”) when migrants began showing up in NYC by the busload to claim their right to taxpayer-funded housing.

What about space-based sunshades funded by the richest and most densely packed cities that are aimed at cooling off only folks in those cities and only for the hottest months of the year? Grok says that New York City could be shaded for less than $5 billion. If we assume that 1 million NYC residents (out of 8.5 million) pay non-negligible taxes, that’s $5,000 per taxpaying New Yorker, essentially nothing. It’s also nothing compared to the NYC government’s $115 annual budget. Maybe Grok is wrong about the likely cost? Or Grok is budgeting for 20 years of improvements in technology?

Unlike air conditioning, a solar shade for the elite has the advantage that it will also cool off non-elite regions of the Earth, at least to a small extent. The constellation of sun shades for NYC could move to provide relief to equatorial cities in the spring and fall and then be repositioned to provide shade to Southern Hemisphere cities during their summer. Maybe a city could still hit 95 degrees due to being near a warm air mass, but there wouldn’t simultaneously be a blazing hot sun. Grok says we can get a 3.5-7-degree F local temperature reduction.

Where else could this work? I.e., what other cities tend to be too hot for some parts of the year, are densely populated, and are rich? Singapore, Shanghai, Hong Kong, Tel Aviv (cue the Jewish Space Lasers!), Dubai, Abu Dhabi, Miami, Doha, Tokyo?

Here’s Grok’s answer after I asked about shading all of Florida and then just NYC:

Technical FeasibilityBuilding a space-based sunshade for New York City (NYC, approximately 800 km² land area) that deploys seasonally from June 1 to September 15 is theoretically possible but would require significant advancements in solar sail technology and orbital mechanics. Unlike global sunshades at the Sun-Earth L1 point (1.5 million km away), a city-specific system would use a constellation of smaller, lightweight reflective sails or mirrors in Earth orbit to provide targeted shading. This draws from early concepts like Hermann Oberth’s 1920s-1970s proposals for orbiting mirrors (100-300 km diameter) to weaken solar radiation over specific regions.

Modern adaptations could leverage “statite” configurations—solar sails that balance solar radiation pressure against gravity to “hover” over fixed points, such as high latitudes.

Design and Deployment: A constellation of 10-50 sails (each 10-50 km², depending on orbital altitude) could achieve 20-50% insolation reduction over NYC during summer daylight hours. Sails would use ultra-thin materials (e.g., polymeric films with SiO₂ nanotubes, areal density ~1-4 g/m²) for reflectivity. For seasonal use, sails could be oriented edge-on to the Sun (non-reflective) outside summer, or maneuvered into storage orbits using onboard ion thrusters or solar pressure. Deployment might involve launching folded sails from Earth, unfolding in orbit, and positioning via solar sailing—reaching operational orbits in weeks to months.

Orbital Configurations: To maintain persistent shading over NYC (40.7°N), sails could use inclined geosynchronous orbits (24-hour period) or Molniya-like elliptical orbits to linger overhead during daylight. Statite modes could allow quasi-stationary hovering by tilting sails to counteract gravity, enabling seasonal activation without full repositioning.

Studies on L1 sunshades show orbiting disks can mitigate seasonal temperature variations by ~0.5-1°C globally, suggesting adaptability for targeted summer cooling (e.g., reducing NYC heat by 2-4°C).

Size and Materials: Total sail area might be 500-2,000 km² for effective shading, far smaller than global proposals (3-4 million km²). Advanced sails (reflectance >90%) could minimize mass to 1,000-5,000 tons, deployable via reusable rockets like Starship.

Challenges and LimitationsEngineering and Launch: Solar sail tech is nascent (e.g., NASA’s Solar Cruiser prototype); scaling for precision control over a city requires breakthroughs in attitude control and collision avoidance. Launching via 50-100 missions (at $50-100/kg) could take 5-10 years, with risks of orbital debris exacerbating Kessler syndrome.
Cost: Estimated $500 million to $5 billion, based on scaled-down global figures ($5-10 trillion for planetary-scale).

This includes R&D ($100-500 million), launches, and maintenance. Seasonal operations add complexity, potentially increasing costs by 20-30% for repositioning fuel/thrusters.
Environmental and Ethical Issues: Targeted shading could alter local weather (e.g., reduced rainfall or photosynthesis impacts), affect solar energy production, or cause uneven cooling in adjacent areas. Global governance is needed, as unilateral deployment raises international concerns. Sudden failure (e.g., solar storm damage) could lead to rapid warming.

Current Status: No prototypes exist for regional shading; efforts focus on global concepts via groups like the Planetary Sunshade Foundation.

Implementation might be viable in 20-30 years with investments in reusable launchers and in-space manufacturing.

In summary, while conceptually sound and more feasible than Florida-scale due to NYC’s smaller size, this remains speculative geoengineering. Ground-based alternatives (e.g., cool roofs, urban trees) offer quicker, cheaper relief.

Sustained R&D could make it a complementary tool for urban heat mitigation.

Related:

Full post, including comments

An importer’s view of the tariff situation

by philg

From a friend who imports stuff from Europe…

As we all sit and think about the tariff decision today, let us all remember why this is being litigated in the first place. Anyone who has read a contract knows that the first part quite often contains definitions.

The law that is the Administration’s basis for the tariffs said that the President could “regulate” trade in certain circumstances. “Regulate” was not defined. So most of this was about whether the word “regulate” included tariffs.

All of this could have been avoided if any of the mediocre (or worse) lawyers in Congress had defined the word “regulate.” Instead billions of dollars of tariffs have been collected and industry has been put in turmoil because the idiotic mediocre lawyers in Congress could not define their terms.

So the next time you ask for these utter fools to pass a law to save the country, please remember this moment.

Full post, including comments

Should Tesla offer an Osborne effect lease?

by philg

A fair number of my friends are Tesla owners. Most of them live in the Northeast or California so cue the apologetic stickers, e.g.,

Existing owners are a car company’s best customers, typically accounting for more than 50 percent of purchases. Tesla has killed upgrade demand among my friends by announcing HW5/AI5 in June 2024. At the time, Tesla promised delivery in late 2025, but now they’re saying that AI5 has slipped at least one year to “late 2026”. This is the classic company-killing Osborne effect.

I wonder if Tesla could undo some of this self-inflicted damage by offering an “Osborne effect lease”. A consumer would sign up today for a 3-year lease but if AI5 makes it out the door within those three years the consumer is guaranteed either (a) a retrofit of AI5 to his/her/zir/their car, or (b) a swap-out of the HW4 car for an AI5 car within 3 months of AI5 mass production and at a price that is set today, e.g., $200/month extra (some of the extra money going toward the wonderfulness of AI5 and some because the swapped-in car is newer than the originally leased car).

Separately, my friend with a 2025 Tesla Y made it from southern New Jersey to Cambridge, Maskachusetts the other day via FSD and touched the steering wheel only twice. Once was to park at a charger and once was to park in Harvard Square. Given the chaotic nature of road markings in Cambridge and in NYC this seems like more of an achievement than my neighbor’s regular Gainesville-Jupiter drives on FSD.

Full post, including comments

How does the Gaza Health Ministry operate with no food, building, electricity, or Internet?

by philg

Today’s question is how the Gaza Health Ministry operates. This organization is as authoritative for progressives today as Anthony Fauci was during coronapanic.

We are informed by CNN that Gazans are starving. Maybe if they walk 10 miles they can get a few scraps, but probably they will be shot by Israeli “snipers” if they approach the feeding stations. We are informed by the New York Times and CNN that all buildings in Gaza have been destroyed. We are informed that Gaza has no electricity or Internet. So a group of people (the “Ministry”) with no food, no building, and no Internet manages to work together all day every day and produce reliable precise-to-six-digits statistics and press releases. Here’s a recent example from taxpayer-funded PBS… “Over 60,000 Palestinians have died in the Israel-Hamas war, Gaza’s Health Ministry says”:

Over 60,000 Palestinians have been killed in the 21-month Israel-Hamas war, Gaza’s Health Ministry said Tuesday. The ministry, part of the Hamas-run government, said the death toll has climbed to to 60,034, with another 145,870 people wounded since Hamas’ Oct. 7, 2023, attack. … An additional 14 Palestinians were killed while attempting to access aid near the American and Israeli-backed Gaza Humanitarian Fund site in central Gaza, according to local hospitals. … Gaza, which has teetered on the brink of famine for two years

A starving double-chinned woman and a starving man (gender assumed based on wearing of abaya/hijab or not) in a yellow shirt with an ample belly lead off in this August 21, 2025 video from the Guardian:

Underneath the video, the journalists quote the ministry that works without an intact building, food for its employees, or Internet yet gathers precise statistics:

On Sunday, the Gaza health ministry said eight more people had died of malnutrition and starvation, raising deaths from such causes to 289 people, including 115 children, since the war started.

From the Islamic Republic of Minneapolis, “Gaza Health Ministry says more than 63,000 Palestinians killed as Israel-Hamas war drags on” (Star Tribune, August 29, via AP):

Gaza Health Ministry says more than 63,000 Palestinians killed as Israel-Hamas war drags on.

But they don’t say where this ministry is located or where its press releases are to be found?

Full post, including comments

Requiem for a 20-year-old Netgear switch

by philg

From exactly 20 years ago:

This had been powering the hard-wired network in the walls of my old Harvard Square apartment ($125/run in pre-Biden dollars, facilitated by an attic above) and failed at some point in 2025. The front of the device was showing only one link light and it wasn’t on a port to which anything was connected. I found to my delight that the GS116 was still made by Netgear(!) and assumed that the new one ($100 for unmanaged) would have the same mounting pattern and, therefore, I could slide it up underneath the desk using the same screws. Sadly, however, they changed the spacing on the mounting holes.

I think that we need to give a shout-out to the Netgear engineers circa 2000 who made a machine that could run almost forever in Internet terms.

(Why didn’t I upgrade to a 10G switch? Xfinity is the monopoly Internet provider in most of Cambridge, Maskachusetts (the officials who wisely run the world’s smartest city rejected Verizon FiOS some years ago) and one is lucky to get 200 Mbits download throughput on their “gigabit” service, which is throttled to 35 Mbits upload.)

Full post, including comments

Who has experienced a USB-C jack failure due to dirt/lint?

by philg

My 9-month-old pocket-carried iPhone 16 Pro Max stopped connecting reliably to USB-C cables. This had never happened in the bad old Lightning jack days. There’s an Apple Store 15 minutes south of us in Palm Beach Gardens (home to Tiger Woods’s former plaintiff/club-wielding attacker). The “geniuses” there cleaned it out in their back room so I can’t say what tools they used.

I had thought that the EU-mandated euthanasia for Lightning was a pure boon to consumers, but maybe others have noticed the same fragility?

Here’s what ChatGPT has to say:

Yes — USB-C jacks are generally more susceptible to dirt and lint issues than Lightning, though both can fail from pocket debris. Here’s why:

1. Connector Design

  • USB-C: The jack is a receptacle with 24 tiny exposed contacts along the walls. Debris can sit in the bottom of the port and prevent the plug from seating fully, or it can block some pins. Because the plug’s tongue is solid and must slide into the port, any pocket lint at the bottom tends to get compacted and cause “won’t charge” problems.
  • Lightning: Apple’s Lightning jack has fewer contacts (8), located only on the plug itself. The port on the device is a simple cavity with spring contacts on the sides. Dirt buildup still happens, but there’s less opportunity for debris to block a critical central “tongue.”

2. Durability in Daily Use

  • USB-C: More fragile if debris prevents a flush fit — wiggling or forcing a plug against lint can bend pins inside the port.
  • Lightning: More forgiving; the plug’s flat contacts scrape against the spring pins, which can tolerate a little more obstruction.

3. Real-World Failure Patterns

  • Many phone repair shops report that the number one cause of “dead” USB-C ports is lint compaction, especially in phones carried in jeans pockets. The plug simply won’t go in all the way until the port is cleaned with a pick or compressed air.
  • Lightning ports can also clog, but they usually fail from mechanical wear on the spring contacts after many years, rather than lint preventing insertion.

4. Cleaning and Maintenance

  • USB-C: Needs more frequent cleaning if the device is often pocket-carried. A wooden or plastic pick works best.
  • Lightning: Easier to clean, less prone to deep compaction.

🔹 Best Practices for Cleaning a USB-C Jack

1. Power Down First

  • Turn off the device completely.
  • If possible, unplug the battery (for laptops) to eliminate any chance of shorting pins.

2. Start With Compressed Air

  • Use a can of compressed air held upright.
  • Give the port short, controlled bursts at different angles.
  • Avoid holding the nozzle too close or using “liquid spray” mode (which can cause condensation).

3. Use a Non-Conductive Pick if Needed

If air alone doesn’t work:

  • Use a wooden or plastic toothpick or a specialized port-cleaning tool.
  • Very gently scrape or lift out compacted lint at the bottom of the port.
  • Work slowly to avoid scratching or bending the pins on the side walls.

4. Optional: Isopropyl Alcohol for Sticky Residue

  • If there’s grime (e.g., soda, lotion), lightly dampen a lint-free swab with 99% isopropyl alcohol.
  • Gently wipe the inside walls.
  • Let it dry completely before powering on.

5. What NOT to Do

  • ❌ Don’t use metal objects like paperclips or needles — they can short pins or bend contacts.
  • ❌ Don’t soak the port in liquid cleaner.
  • ❌ Don’t blast with high-pressure air compressors (too strong, may damage).

🔹 Maintenance Tips

  • Keep devices in a case that shields the port from pocket lint.
  • Store with a dust plug (tiny rubber caps for USB-C are cheap).
  • Clean ports periodically if you carry your phone in jeans pockets.

Should folks who do all of their charging and CarPlaying, etc. wirelessly invest 10 cents in a dust cover?

Full post, including comments

Amtrak Acela ride from New York to Boston

by philg

Amtrak is spending $2.5 billion of your tax dollars on “NextGen” Acela train sets. Here’s a report on an August 2025 trip via OldGen Acela.

My $275 first class ticket from New York to Boston entitled me to use the Metropolitan Lounge at New York’s Penn Station, now named after Senator Daniel Patrick Moynihan, a liberal in his time and a reactionary conservative by the standards of today’s Democratic Party. In his 1965 report, for example, he seems to be engaging in victim-blaming:

The percent of nonwhite families headed by a female is more than double the percent for whites. Fatherless nonwhite families increased by a sixth between 1950 and 1960, but held constant for white families. It has been estimated that only a minority of Negro children reach the age of 18 having lived all their lives with both of their parents. …The Breakdown of the Negro Family Has Led to a Startling Increase in Welfare Dependency.

(Little did he imagine that white Americans would be incented to catch up to their Black brothers, sisters, and binary-resisters by the Family Support Act of 1988, which required states to set up child support formulae such that family court lawsuits would have a guaranteed minimum profit.)

Here’s the view from the lounge’s balcony:

The interior of the lounge, which offers free sandwiches, salads, chips, non-alcoholic drinks, etc.:

Due to the lack of crowding and the friendliness of the employees, this lounge is 20X nicer than what the airline lounges have become. Working WiFi was provided for those who want to sit with their laptops.

Separately, the lounge has a good view of the main hall’s video wall in which the New York City government promises to hire, and then retain forever (“strong labor protections”), anyone fired by the federal government for laziness and/or incompetence:

My initial Facebook post about the trip:

Just leaving Penn Station on time via Acela. Great crew on AMTRAK. The train itself bucks and bumps so much that they should have announcements in Mandarin to reassure our Chinese brothers and sisters that a derailment isn’t imminent (high-speed rail in China is perfectly smooth even above 200 mph). Maybe the idea is that everyone will be too plastered to notice?

The food in the lounge, though simpler, was generally superior.

Some hard-working guys:

My follow-up comments to the Facebook post, not in quote style for readability…

Cars and trucks were blowing past us on I-95, but now I think we’re matching their speed.

We’re in the middle of a one-hour run to New Haven. The iPhone reports a blistering speed.

we’ve cut back to a speed where we won’t have to adjust our watches for time dilation under Mileva Marić’s theory of special relativity (popularized by her husband).

Wifi doesn’t work. AMTRAK needs to adopt Starlink! [The new $2.5 billion cars won’t have Starlink, but “5G” (same idea as what the old cars have and that didn’t work for my trip.]

Our AI overlords have access to data showing that I’m on a train track and moving at 30-60 mph. Also it is time to point out that an AirTag is moving with me on the same track. Maybe this would be a useful warning for someone who owned a private rail car attached to the back of an AMTRAK train?

Approaching Westerly, CT at left lane pickup truck speed on I-95 in South Florida.

15 minutes before Providence we hit 150 mph:

Arrived on time at South Station, just under four hours after leaving Penn Station. Perfect no-waiting connection to the Red Line. Got to the Harvard Square apartment at almost exactly the same as Google Maps had predicted for leaving Penn Station at 6:30 pm by car.

Full post, including comments

Credentialed expert consensus: SpaceX Starship will never work

by philg

Two months ago, New York Magazine Intelligencer (“intelligence” even in the name of the publication), “Is Elon Musk’s Starship Doomed? The future of SpaceX keeps blowing up, and no one knows if he can fix it.”:

“Having a rocket ascend a few hundred meters and blow up is not a success to me,” says Dallas Kasaboski, an analyst who covers the space industry for the research firm Analysys Mason.

Reusing both parts of a rocket, instead of just one, sounds like a modestly more ambitious undertaking — if you can do it once, why not twice? — but in fact, it’s exponentially harder, since the amount of energy to be dissipated while returning to Earth goes up not as a linear function of velocity but its square. The energy gets turned into heat, which sends the temperature of the Starship soaring into the thousands of degrees, such that it glows white-hot. To reduce the effects of this heat, the rocket’s exterior is coated in thermal tiles. But these are heavy. So is the internal bracing required to hold the ship together during the turbulence of reentry.

For critics, like Substack writer Will Lockett, the fact that Starship has failed so many times in a row is proof that the concept is fundamentally unworkable. “SpaceX is having to make the rockets too light, resulting in them being fragile, meaning that just the vibrations from operation with a fraction of its expected payload would be enough to destroy the rocket,” he wrote in one typically acerbic post.

“I remember reading a book in the 1970s about how the space shuttle would be flying every two weeks and cost low hundreds of dollars per pound to orbit,” says Grant Anderson, co-founder of the space-systems-maker Paragon Space Development Corporation, which helped launch SpaceX in 2002. “It never did that, obviously.” … When would we know if Starship is toast? One major red flag is persistent failures. “If you have a failure at the same stage twice for the same reason, and they can’t solve the problem, that’s an indication that there’s a design flaw that’s more than just an ‘Oops!’” says Anderson.

Readers: What do you think? Is it possible that Elon Musk and SpaceX are right and the expert consensus is wrong? Is there any way that Starship could one day make it into orbit without exploding? And, perhaps one day many years or decades from now, that both booster and Starship could land back on Earth?

Full post, including comments

Our AI overlords will need more power than the Sun generates within 43 years

by philg

My question to Grok: “Extrapolating from the past three years how long before running LLMs consumes more power than the entire Sun generates?”

Extrapolating from the growth in global AI data center power demand over the past three years (2022–2025), which serves as a proxy for LLM power consumption given that LLMs drive much of the recent AI surge, we can estimate the timeline. Data from SemiAnalysis (as cited in a RAND report) shows the following facility power demand for AI data centers:

  • 2022: 1.451 GW
  • 2023: 4.28 GW
  • 2024: 10.6 GW
  • 2025: 19.9 GW

To arrive at the extrapolation:

  1. Fit an exponential growth model to these data points. Using linear regression on the natural log of power vs. year (with 2022 as t=0), the slope is approximately 0.876, corresponding to an annual growth factor of e^0.876 ≈ 2.401 (i.e., power multiplies by about 2.4 each year on average).
  2. The Sun generates 3.826 × 10^26 watts of power, equivalent to 3.826 × 10^17 GW.
  3. Starting from 19.9 GW in 2025, the factor needed to exceed the Sun’s output is (3.826 × 10^17) / 19.9 ≈ 1.924 × 10^16.
  4. Solve for years n: 2.401^n = 1.924 × 10^16, so n = log(1.924 × 10^16) / log(2.401) ≈ 43.

Thus, it would take approximately 43 years from 2025 (i.e., by around 2068). This assumes the exponential growth rate from the past three years continues indefinitely, though in reality, physical, technological, and economic constraints would likely slow or halt such growth long before reaching solar scales.

Note that ChatGPT 5 failed to answer this question, identically worded. It used data center power consumption growth, not LLM- or AI-specific power consumption growth.

When does the exponential growth stop? We can power a human brain with an Egg McMuffin. Is the answer an analog chip with a digital interface? IEEE has a 2024 article on a Chinese design using optical computing that supposedly cuts power consumption by 99%. But if these improvements were real would the AI companies be signing 20 year leases for nuclear power?

Full post, including comments