Financial Nonsense
Sign in<br>Subscribe
Many years ago, in an earlier era of silliness about a looming ed-tech revolution, a startup founder sent me an email, asking that I please hold off from writing anything about the company. They'd experienced a huge surge in demand for their app – it was fall, students were headed back-to-school. He confessed that, until their new round of funding actually cleared the bank, they were struggling to pay their bills, particularly a very large bill for Twilio, the messaging service that was, in fact, almost the entirety of the functionality of what they were offering to schools. But unlike Twilio, this particular app was "free" – I put that word in quotation marks because, of course, nothing is. There are real costs to running a business: paychecks have to be cut, operational expenses have to be paid for. What enabled this specific startup, and so many like it, to offer their product at no charge was a subsidy of sorts, in the form of venture capital, which has promoted a new, strange business model, allowing companies to avoid thinking about revenue let alone profitability – ostensibly just in the early stages, but increasingly (and dangerously), long into the lifespan of a company.<br>Schools love "free," of course, as they're almost always operating with budgetary constraints (and/or with political threats to axe even more of their funding). So "free" software has been able to make substantial in-roads into classrooms -- particularly alongside the narrative about the compulsory usage of ed-tech -- and many educators and administrators have been able to convince themselves that there really are no costs to adoption.<br>But at some point, the venture capital runs out. At some point, the bill comes due. At some point, a startup has to determine what part of the product or service it will charge for. "Find the thing that schools can't do without," I heard one founder quip at an ed-tech event years ago, "and make them pay for that." "Make it painful for them to stop," he added, suggesting that students with disabilities made for a very good target for this strategy.<br>Even when these pricing practices aren't nefarious (although frankly they mostly are -- that's capitalism), we all understand how software can be "painful to stop" -- not because we're "addicted" (I will always hate that framwork) or because of some weird "brand loyalty," but simply because we have used a particular tool for so long; we are familiar with its functionality, and we don't want to have to learn a new interface; and, of course, these apps have so much of our work, our data, and tech companies have made it virtually impossible to move any of it elsewhere.<br>There are other problems with "free" ed-tech too. As it costs nothing to adopt (and then likely reject) different apps, schools have found themselves using thousands of different apps (an average of almost 2600 according to a report a few years ago -- that report found that students accessed about fifty unique applications over the course of the year). Student and teacher data is strewn across these tools, often with little regard for privacy or security, let alone any semblance of curricular consistency or a shared educational experience.<br>This flurry of tool adoption is typically (mis-)read as enthusiasm for ed-tech rather than what's far more likely: most school software is shit; you download it, use it once or twice, and yikes, you never touch it again. And even when there is enthusiasm -- such seems to be the case with students' usage of ChatGPT to do their homework -- we should still pause before celebrating this as some educational marvel. We should still pause before interpreting any of this as indication there's some insatiable demand for ed-tech or for generative AI because it's good. Mostly, it's because it's free.<br>As Dave Karpf wrote back in May, "we are still in the free-trial-period of most AI products. Everything is being subsidized. That renders a false picture of the demand curve." "Free" distorts demand; "free" also distorts a business's viability. AI companies -- much like many ed-tech startups who like to boast, come August, an explosion in sign-ups for their free app -- might be able to show off an incredible growth in the number of subscriptions; but that growth is actually driving their revenue down not up.<br>(And now, as AI companies are starting to charge based on token-usage not simply a flat subscription, many of their customers are shocked at the bill. "Companies are scrambling to stop spending so much on AI," as 404 Media reports.)<br>What we are witnessing with AI is a demand for compute, the AI companies' demand for compute: their demand for data centers, their demand for data storage, their demand for data processing. Generative AI seems to have an insatiable demand for power (literally and figuratively) and for resources; and the latter isn't just about water -- the industry has also been gobbling up the components necessary to...