Samsung & Google – Engineering Disease pt. V

Samsung is still not quite cured.

• Samsung’s deal with Perplexity will see Samsung’s in-house products become better, but I am not sure how much anyone will notice given that users buy Android for Google rather than Samsung.
• Perplexity is primarily a search engine that uses natural language and its knowledge graph from crawling the Internet rather than a large language model that has been trained on the knowledge of the Internet.
• The result is that its answers can lay claim to greater accuracy than LLMs, which have a tendency to hallucinate, but its ability to cross-reference or generate text and images is more limited.
• Until recently, Perplexity had a substantial edge in my view, but with the addition of features such as Deep Research, Google and OpenAI have materially closed that gap.
• Perplexity’s pricing for output tokens via its API also tends to suggest that it is cheaper for Perplexity to process its requests than it is for an LLM, especially those that do many rounds of “reasoning”.
• Hence, I think that Perplexity occupies a unique space in the market, but with the amounts that the LLM crowd is spending on improving their products and driving new usage, its position may become somewhat precarious.
• This leads us to Samsung, as Perplexity is raising $500m at a $14bn valuation, where it looks like Samsung will be one of the larger investors.
• As part of the investment, it looks like Samsung will get access to Perplexity and by infusing Samsung services such as the execrable Bixby and Samsung’s browser with Perplexity’s capabilities, they will become a lot better than they are today.
• Samsung and Perplexity have been in discussions for some time over a variety of topics like an AI-powered operating system and AI agents but I suspect everything hinges on whether Samsung makes the investment.
• While Perplexity will undoubtedly make Samsung’s in-house services better, the key question is whether anyone will notice and whether Samsung should be making its own services at all.
• Samsung’s services are fairly widely regarded as bloatware and a reason not to buy a Samsung device, while Bixby is the source of endless online jokes (see here).
• I have long been of the opinion that Samsung’s obsession with creating software (that is inferior to Google’s) and Google’s obsession with making hardware (that is inferior to Samsung’s) are classic cases of engineering disease.
• Engineering disease is a malady that often infects companies with large cash balances, where they start making products because they can and not because they should from a business perspective.
• Both Samsung and Google have been making good progress towards finding a cure, as the collaboration on Android XR demonstrates.
• Both are working together to create a single hardware and software platform for the Metaverse, and while Google is making a pair of glasses, I think it will be Samsung that makes all the running in terms of shipments.
• Adding Perplexity to its services that compete with Chrome and Gemini is clearly a step in the wrong direction, but the good news is that I am not convinced that it will matter very much.
• This is because usage of Samsung apps that are alternatives to Google’s ecosystem is very low in my opinion, and so I am not convinced that anyone will notice the upgrades.
• The real winner is Perplexity as it will be able to use Samsung as an anchor client, and if no one uses its services on Samsung devices, it won’t have to swallow much in the way of compute cost.
• This is a slip by Samsung, but in general, it has made progress on focusing on what it is good at and staying away from things that its partners can best create.
• If it can make progress on fixing its issues in high bandwidth memory and qualify its HMB4 data centre product with Nvidia, then we should see a recovery in the valuation of the shares also.
• I think that this is likely, which is why I continue to hold a position in Samsung.