{"id":479000,"date":"2026-05-11T10:20:11","date_gmt":"2026-05-11T10:20:11","guid":{"rendered":"https:\/\/www.europesays.com\/ie\/479000\/"},"modified":"2026-05-11T10:20:11","modified_gmt":"2026-05-11T10:20:11","slug":"cuda-proves-nvidia-is-a-software-company","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ie\/479000\/","title":{"rendered":"CUDA Proves Nvidia Is a Software Company"},"content":{"rendered":"

Forgive me for starting with a clich\u00e9, a piece of finance jargon that has recently slipped into the tech lexicon, but I\u2019m afraid I must talk about \u201cmoats.\u201d Popularized decades ago by Warren Buffett to refer to a company\u2019s competitive advantage, the word found its way into Silicon Valley pitch decks when a memo purportedly leaked from Google<\/a>, titled \u201cWe Have No Moat, and Neither Does OpenAI,\u201d fretted that open-source AI would pillage Big Tech\u2019s castle.<\/p>\n

A few years on, the castle walls remain safe. Apart from a brief bout of panic when DeepSeek<\/a> first appeared, open-source AI models have not vastly outperformed proprietary models. Still, none of the frontier labs\u2014OpenAI, Anthropic, Google\u2014has a moat to speak of.<\/p>\n

The company that does have a moat is Nvidia. CEO Jensen Huang has called it his most precious \u201ctreasure.\u201d It is not, as you might assume for a chip company<\/a>, a piece of hardware. It\u2019s something called CUDA. What sounds like a chemical compound banned by the FDA may be the one true moat in AI.<\/p>\n

CUDA technically stands for Compute Unified Device Architecture, but much like laser or scuba, no one bothers to expand the acronym; we just say \u201cKOO-duh.\u201d So what is this all-important treasure good for? If forced to give a one-word answer: parallelization.<\/p>\n

Here\u2019s a simple example. Let\u2019s say we task a machine with filling out a 9\u00d79 multiplication table. Using a computer with a single core, all 81 operations are executed dutifully one by one. But a GPU with nine cores can assign tasks so that each core takes a different column\u2014one from 1\u00d71 to 1\u00d79, another from 2\u00d71 to 2\u00d79, and so on\u2014for a ninefold speed gain. Modern GPUs can be even cleverer. For example, if programmed to recognize commutativity\u20147\u00d79 = 9\u00d77\u2014they can avoid duplicate work, reducing 81 operations to 45, nearly halving the workload. When a single training run costs a hundred million dollars, every optimization counts.<\/p>\n

Nvidia\u2019s GPUs were originally built to render graphics for video games. In the early 2000s, a Stanford PhD student named Ian Buck, who first got into GPUs as a gamer, realized their architecture could be repurposed for general high-performance computing. He created a programming language called Brook, was hired by Nvidia, and, with John Nickolls, led the development of CUDA. If AI ushers in the age of a permanent white-collar underclass and autonomous weapons, just know that it would all be because someone somewhere playing Doom thought a demon\u2019s scrotum should jiggle at 60 frames per second.<\/p>\n

CUDA is not a programming language in itself but a \u201cplatform.\u201d I use that weasel word because, not unlike how The New York Times is a newspaper that\u2019s also a gaming company, CUDA has, over the years, become a nested bundle of software libraries for AI. Each function shaves nanoseconds off single mathematical operations\u2014added up, they make GPUs, in industry parlance, go brrr.<\/p>\n

A modern graphics card is not just a circuit board crammed with chips and memory and fans. It\u2019s an elaborate confection of cache hierarchies and specialized units called \u201ctensor cores\u201d and \u201cstreaming multiprocessors.\u201d In that sense, what chip companies sell is like a professional kitchen, and more cores are akin to more grilling stations. But even a kitchen with 30 grilling stations won\u2019t run any faster without a capable head chef deftly assigning tasks\u2014as CUDA does for GPU cores.<\/p>\n

To extend the metaphor, hand-tuned CUDA libraries optimized for one matrix operation are the equivalent of kitchen tools designed for a single job and nothing more\u2014a cherry pitter, a shrimp deveiner\u2014which are indulgences for home cooks but not if you have 10,000 shrimp guts to yank out. Which brings us back to DeepSeek. Its engineers went below this already deep layer of abstraction to work\u00a0directly in PTX, a kind of assembly language for Nvidia GPUs. Let\u2019s say the task is peeling garlic. An unoptimized GPU would go: \u201cPeel the skin with your fingernails.\u201d CUDA can instruct: \u201cSmash the clove with the flat of a knife.\u201d PTX lets you dictate every sub-instruction: \u201cLift the blade 2.35 inches above the cutting board, make it parallel to the clove\u2019s equator, and strike downward with your palm at a force of 36.2 newtons.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"Forgive me for starting with a clich\u00e9, a piece of finance jargon that has recently slipped into the…\n","protected":false},"author":2,"featured_media":479001,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[74],"tags":[289,19434,5149,18,19,17,209773,292,37216,22152,864,82],"class_list":{"0":"post-479000","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-technology","8":"tag-artificial-intelligence","9":"tag-coding","10":"tag-computers","11":"tag-eire","12":"tag-ie","13":"tag-ireland","14":"tag-machine-readable","15":"tag-nvidia","16":"tag-platforms","17":"tag-programming","18":"tag-software","19":"tag-technology"},"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@ie\/116555443532981922","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/479000","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/comments?post=479000"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/479000\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media\/479001"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media?parent=479000"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/categories?post=479000"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/tags?post=479000"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}