{
  "$type": "site.standard.document",
  "content": "---\ntitle: \"DeepSeek and Shallow Moats: Implications for Higher Education\"\ndescription: \"Universities betting big on exclusive AI partnerships risk repeating the MOOC\n  mistake. DeepSeek R1 shows why staying provider-agnostic is the smarter play.\"\n---\n\n> This piece originally published in Times Higher Education\n\n<!-- DeepSeek's arrival may have spooked the markets, but what does it\nmean for the research and development of LLMs? Higher education should avoid\nputting all its eggs in one GenAI basket, writes Ben Swift -->\n\nThe higher education sector has form for betting on technological moats that\nturned out to be mirages. In the early 2010s we rushed to build MOOC platforms,\nconvinced we'd need our own infrastructure to survive in the digital age. A\ndecade later, most of those courses---and the custom platforms we built to host\nthem---have been abandoned. As AI reshapes education, we risk repeating this\ncostly mistake.\n\n[DeepSeek R1](https://github.com/deepseek-ai/DeepSeek-R1) recently made a\nsplash---you may have heard of it (or even\n[downloaded it](https://techcrunch.com/2025/01/27/deepseek-displaces-chatgpt-as-the-app-stores-top-app/)).\nIt's technically impressive,\n[meeting or beating similar models from OpenAI on many benchmarks](https://arxiv.org/pdf/2501.12948),\nand it's also [\"open weight\"](https://epoch.ai/blog/open-models-report), so\nanyone can download and run the model if they have the hardware to run it.\n\nDeepSeek didn't even just release one model---they released\n[a few different models](https://arcprize.org/blog/r1-zero-r1-results-analysis)\n(each with slightly different tradeoffs). They also describe in their research\npaper how they trained them all for a significantly lower cost, in terms of time\nand money, than their competitors. The research community needs more time to\nevalutate these claims, but it does seem like this could be at least a small\nbreakthrough in reducing the amount of resources it takes to train a new LLM.\n\nThe market certainly seemed to think so: Nvidia's shares lost nearly \\$US600bn\nin one day, based on fears that customers wouldn't need to buy as many of their\nAI accelerator chips to train their models. Although even that story's\ncomplicated---DeepSeek is a \"reasoning\" model, which trades off less time and\nresources for training with more time and resources for inference (i.e. running\nthe model to generate text).\n\nStill, DeepSeek's technical achievements are impressive, but the deeper story is\nwhat it tells us about the state of LLM research and development. The argument\nof the\n[leaked 2023 Google memo](https://semianalysis.com/2023/05/04/google-we-have-no-moat-and-neither/)\nasserting \"we have no moat, and neither does OpenAI\" seems to be holding up.\nDespite trillions of dollars of investment, it really does still seem like an\nupstart can come out of nowhere to release (and share) something that's\ncompetitive with state of the art offerings from the tech giants. This poses a\nstrategic question for higher education leaders: how should institutions\nposition themselves in response?\n\nSome universities have already placed significant bets, signing exclusive\npartnerships with major AI companies. The California State University system's\ndeal with OpenAI will\n[provide ChatGPT access to 500,000 students and faculty](https://openai.com/index/openai-and-the-csu-system/).\nUNSW Sydney has\n[inked a similar agreement (albeit on a smaller scale)](https://www.unsw.edu.au/newsroom/news/2024/12/UNSW-Sydney-signs-landmark-agreement-with-OpenAI).\nThese moves reflect an understandable desire to get ahead of the curve, but they\nmay also lock institutions into particular tools and ecosystems at a time where\nnew, and perhaps better, alternatives are emerging. The higher education sector\nis\n[facing huge financial challenges](https://universitiesaustralia.edu.au/wp-content/uploads/2024/11/UA091-Critical-challenges-in-Australias-university-sector_v2.pdf),\nand these contracts take precious resources away from faculty salaries or tutor\nbudgets or any of the other functions of the institution.\n\nThe emergence of models like DeepSeek R1 is a timely reminder that things in AI\nare still moving fast. Rather than pursuing exclusive relationships with\nspecific providers, institutions might better serve their communities by staying\nprovider-agnostic (where they engage at all). This approach acknowledges both\nthe rapid pace of technical change and the likelihood that tomorrow's leading\nmodels may come from unexpected sources.\n\nFor individual educators, DeepSeek's release reinforces what many of us have\nalready realised: the specific model matters less than how we integrate AI\ncapabilities into our pedagogical practice. Whether students use GPT-4, Claude\nor DeepSeek (and let's face it, they _will_) the fundamental challenges remain\nthe same. How do we design assessments that meaningfully evaluate learning in an\nAI-augmented world? How do we help students develop the critical thinking skills\nto effectively collaborate with AI tools?\n\nFor university administrators and planners, these developments suggest a few key\nprinciples:\n\n1. avoid long-term exclusive commitments to specific AI platforms or providers\n2. invest in developing institutional AI literacy and governance frameworks\n3. focus on building adaptable infrastructure that can accommodate multiple AI\n   tools\n\nThere may be other upsides to these developments. As model training and deployment\ncosts decrease, universities may find it increasingly feasible to develop\nspecialised models for specific academic domains or research applications. Such\nprojects could focus on specific institutional needs rather than attempting to\ncompete with general-purpose models.\n\nWhat DeepSeek R1 really shows is that in the AI era, competitive advantage will\ncome not from controlling access to certain models, but from skillfully\nintegrating AI capabilities into our core educational mission. Universities that\nbuild their strategies around particular AI platforms risk finding themselves\ntrapped in technological dead ends, while those that focus on developing\ninstitutional AI literacy and adaptable frameworks will be better positioned to\nembrace whatever technological developments emerge.\n\nThe real moat in higher education isn't technological; it never has been. It's\nthe ability to teach well and generate new knowledge. Technology is merely a\ntool---in reality, a _system_ of tools, people and other resources---in service\nof these fundamental goals. The winners won't be those who bet early on the\nright AI platform, but those who most effectively help their communities master\nthe art of learning and creating in an AI-augmented world.\n",
  "createdAt": "2026-05-13T23:14:44.779Z",
  "description": "Universities betting big on exclusive AI partnerships risk repeating the MOOC mistake. DeepSeek R1 shows why staying provider-agnostic is the smarter play.",
  "path": "/blog/2025/02/18/deepseek-and-shallow-moats",
  "publishedAt": "2025-02-18T00:00:00.000Z",
  "site": "at://did:plc:tevykrhi4kibtsipzci76d76/site.standard.publication/self",
  "textContent": "Universities betting big on exclusive AI partnerships risk repeating the MOOC mistake. DeepSeek R1 shows why staying provider-agnostic is the smarter play.",
  "title": "DeepSeek and Shallow Moats: Implications for Higher Education"
}