Hacker News Reader

Comment by nostrademons | Hacker News Reader

Yup. The largest ones could afford to buy a mid-size country. (AMZN revenue = $386B, AAPL revenue = $274B, GOOG revenue = $182B, Norway GDP = $366B, Vietnam GDP = $340B, Chile GDP = $245B).

Note also that SpaceX has more accurate ballistic missiles than the U.S. and Russia (PeaceKeeper CEP = 40M, while SpaceX routinely lands rockets upright within a ~10M radius on a drone ship), and Google has more data than the CIA & NSA (why else did the NSA try to wiretap them?). The microchip revolution put enormous capabilities in the hands of those organizations that know how to code effectively, such that Silicon Valley startups can often do things that were the sole province of national governments for decades.

View on HN

An ICBM hitting a target purely on internal guidance is an entirely different level of problem from what SpaceX is doing with cooperative positioning systems at the target. Also keep in mind that the "wikipedia numbers" for classified weapons systems are usually "at least" numbers, not exact.

It's not at all clear who has more data between GOOG and NSA. Both don't exactly declare the numbers, but it is known that in several recent years NSA has been the largest single buyer of spinning disks in the world.

Tech companies have indeed become powerful, but let's not get hyperbolic. The world hasn't become shadowrun quite yet.

Let's get hypergoooooolic! What? I've been alone in a room coding for three days.

"in several recent years NSA has been the largest single buyer of spinning disks in the world."

The rest of the world moved to SSDs in ~2010. Google's been storing the whole web on flash since then. (I guess if you're archiving for slow-speed retrieval disks are still useful, but if that's your use-case, why wouldn't you use tape?)

I wasn't really trying to get into some sort of debate about storage technologies, just illustrating that the NSA does have a whole lot of data.

Spinning disk is cost competitive with tape robots these days, while offering much better access performance. There's a few different vendors that supply this sort of setup, or you can build your own backblaze style. The core idea is you build storage servers that can hold more disks than can be active at any given moment. Powering up a drive and doing a seek is a lot faster than a tape robot.

> why wouldn't you use tape

Because there are many orders of latency magnitude between spinning rust and tape.

We still use a lot of spinning disk for precisely this reason - you get roughly 5x the online capacity for the same cost and automation does not whine about wanting the new shiny.

That was why I mentioned SSDs - there are orders of latency magnitude between SSDs and spinning disk.

I guess it fits the NSA's use-case though, where they have ~thousands of analysts running complex queries but no external users, rather than ~billions of external users running well-defined queries.

I left the storage industry about 5 years ago, but at least as of then the big cloud providers were still buying lots of spinning rust and showed no signs of stopping. These companies store a lot of long tail data which needs to be available but doesn't need super low latency access. It's still much more economical to store such data on disks than in flash.

I’ve worked on infrastructure (design) engineering teams at a very large company that still does their own datacenters. I can confirm that spinners are still very much in use and are not used for archival purposes - that’s what tapes are designed and used for.

Agreed with the sentiment that modern companies are capable, but disagreed on the rocket comparison -- the comparison is difficult at best.

One is travelling at tremendous speed (Mach 5 or something?) and has milliseconds for course-corrections, the aspires to come to a gentle halt and can close some feedback loops over and over again for thousands of milliseconds.

Edit: If you wish to make a weapons comparison, one might note that SpaceX can launch the mother of all MIRVs over and over again. The lower cost-to-orbit changes the economic calculus of rods-from-god, opening a possibility of a conventional-weapons bombardment from the other side of the planet for the first time.

ICBM's also can't use GPS or equivalent navigation satellites, as those will likely be the first to be taken out in a nuclear exchange. They use inertial guidance (https://www.thedrive.com/the-war-zone/30254/this-isnt-a-sci-... ) maybe together with some star navigation system for mid-course guidance.

> One is travelling at tremendous speed (Mach 5 or something?)

Much higher, about mach 20 to 25. They are moving seriously fast.

Oh, you're right -- it has been ages since I've read about things like ABM intercept problems.

That does a great job of placing the accuracy problem in context: That's ~7 km/second, or perhaps a mile per eyeblink.

Surely one would expect to pay at least a few times annual revenue for a country?

That's revenue <=> GDP, and roughly earnings <=> tax revenues. I believe those are equivalent comparisons: revenue and GDP both include all internal transactions, while earnings and tax revenue reflect what's available for management/government to spend at their discretion.

If a tech company were buying the country, market cap would be the appropriate metric, and market caps for those tech companies are in the $2T range, more than enough to pay a few times annual GDP.

Not if you use a community adjusted ebitda valuation.

Reply on news.ycombinator.com