First things first

One of the more revealing aspects of the Dropbox S-1 filing is its disclosure that it saved about $75 million in two years by (largely) migrating its cloud-storage platform off of Amazon Web Services and onto its own infrastructure. While that level of savings is certainly important for a company trying to rein in losses in anticipation of an IPO, I actually think cost might actually be a secondary reason for making such a move.

After all, we've all been watching the prices drop and options for cloud storage increase like clockwork over the past several years. And, as someone on Twitter commented to me after linking to the story above, Dropbox might well be able to achieve the same cost savings using newer AWS storage options such as Glacier and S3 Infrequent Access. Dropbox executives are not living blindfolded on a remote island; they see the same market trends everybody else sees.

That's I would argue that for companies reaching Dropbox's scale, the move away from the cloud is more about control than it is about cost. Go read the Dropbox Tech Blog over the past couple years and see how the company has been building out its infrastructure in order to maximize efficiency and latency. Or read anything Facebook has to say (including this feature story I wrote on it last year) about why it runs its own data centers. What companies at this scale really want is the ability to improve certain aspects of their infrastructure beyond what might be available within the constraints of a cloud provider. We've seen that happen everywhere from chips to storage to networking (both inside and across data centers).

Further, I think it becomes clear to certain companies -- especially companies storing sensitive or important user data -- that blaming their cloud provider for breaches or downtime isn't going to fly for too long. Debating whether Dropbox can do a better job of these things than AWS or Google can almost misses the point (although it probably has the engineering resources to do a damn fine job). Dropbox needs to be able to act on any issues as quickly as possible without waiting on somebody else (which requires controlling and understanding the infrastructure) and its users need to know the company is willing and able to do so.

Finally, it's becoming increasingly common for companies competing against certain cloud providers to support those providers by using their services. Retail companies publicly stating they won't use AWS is the most common example of this, but a company like Dropbox must feel some of the same things as it tries to out cloud-storage-and-collaboration AWS, Google and Microsoft. This is not just a question of putting money in their pockets, but also of giving them any competitive intelligence into how the business is growing or what types of resources Dropbox is consuming.

Is there really a case against a Google monopoly in AI?

I have a lot to say about this WIRED column advocating for some sort of action of Google's impending monopoly in the smart-home space, but my thoughts are hard to to organize other than that I think it's misguided for a handful of reasons. Here they are, in no particular order:

  • I'm not certain Google actually will come to dominate the smart home. Amazon definitely is winning in device sales at the moment, and to the extent that cloud computing services matter in this fight as much as search prowess does, Amazon is still in a great position. (It also partners with companies like Sonos.)
  • Assuming smartphones are a key part of this market Google could monopolize, it's hard to overlook the fact that the iPhone exists.
  • AI is a much broader space than just smart-home gadgets. While Google arguably dominates AI research today thanks to Google Brain and Deepmind, lots of good research happens elsewhere and the open nature of it means Google's competitors aren't left in the dark. They also do a lot of research.
  • There's a whole other side to the AI product discussion, which has to do with selling AI services to other companies. For cloud providers, this is tied pretty tightly to what they're building for their own devices (as noted above) at the moment (in the form of APIs), but it's unclear just how profitable this is for them. But there's also a boatload of money going into the "enterprise AI" space and specialized AI hardware and other things, and it's anybody's guess who'll actually win there. If the answer turns out to be "not Google or AWS," then their actual dominance upstream could be blunted, as well.
  • I'm not sure this would be anything antitrust regulators would care about (and it might actually make the case for regulating a company like Google), but Google is doing some work in areas like biotech and health care that could end up doing a lot of good for the world. Just this week, the was a Nature article about how Google (and deep learning, in general) is helping biology research, Deepmind announced it's taking its kidney-failure prediction algorithms to the U.S. Department of Veterans Affairs and the Google Brain team released a paper advancing a new approach to general-purpose AI systems. One could argue that regulating who wins in a space as relatively insignificant as smart-home devices is a bad reason for potentially stifling research in these more significant areas.

AI and machine learning

Cloud and infrastructure

Sponsor: Replicated

Data and analytics