{"id":1086,"date":"2026-07-10T02:30:24","date_gmt":"2026-07-09T17:30:24","guid":{"rendered":"https:\/\/itexplore.org\/jp\/columns\/ai-cost-surge-microsoft-spacexai-zml-inference\/"},"modified":"2026-07-10T02:30:24","modified_gmt":"2026-07-09T17:30:24","slug":"ai-cost-surge-microsoft-spacexai-zml-inference","status":"publish","type":"post","link":"https:\/\/itexplore.org\/jp\/columns\/ai-cost-surge-microsoft-spacexai-zml-inference\/","title":{"rendered":"AI\u30b3\u30b9\u30c8\u9ad8\u9a30\u3068\u65b0\u30e2\u30c7\u30eb\u767a\u8868\uff1aMicrosoft\u306e\u81ea\u793eAI\u30b7\u30d5\u30c8\u3001SpaceXAI\u306eGrok 4.5\u3001ZML\u306e\u63a8\u8ad6\u9ad8\u901f\u5316"},"content":{"rendered":"<p>\u672c\u65e5\u306e\u6ce8\u76eeAI\u30fb\u30c6\u30c3\u30af\u30cb\u30e5\u30fc\u30b9\u3092\u3001\u5c02\u9580\u7684\u306a\u5206\u6790\u3068\u5171\u306b\u304a\u5c4a\u3051\u3057\u307e\u3059\u3002<\/p>\n<div class=\"wp-block-vk-blocks-alert vk_alert alert alert-warning has-alert-icon\">\n<div class=\"vk_alert_icon\">\n<div class=\"vk_alert_icon_icon\"><i class=\"fa-solid fa-triangle-exclamation\" aria-hidden=\"true\"><\/i><\/div>\n<div class=\"vk_alert_icon_text\"><span>Warning<\/span><\/div>\n<\/div>\n<div class=\"vk_alert_content\">\n<p>\u3053\u306e\u8a18\u4e8b\u306fAI\u306b\u3088\u3063\u3066\u81ea\u52d5\u751f\u6210\u30fb\u5206\u6790\u3055\u308c\u305f\u3082\u306e\u3067\u3059\u3002AI\u306e\u6027\u8cea\u4e0a\u3001\u4e8b\u5b9f\u8aa4\u8a8d\u304c\u542b\u307e\u308c\u308b\u53ef\u80fd\u6027\u304c\u3042\u308b\u305f\u3081\u3001\u91cd\u8981\u306a\u5224\u65ad\u3092\u4e0b\u3059\u969b\u306f\u5fc5\u305a\u30ea\u30f3\u30af\u5148\u306e\u4e00\u6b21\u30bd\u30fc\u30b9\u3092\u3054\u78ba\u8a8d\u304f\u3060\u3055\u3044\u3002<\/p>\n<\/div>\n<\/div>\n<div class=\"wp-block-group\" style=\"margin-top:40px;margin-bottom:40px\">\n<h2 class=\"wp-block-heading\">SpaceXAI\u304c\u4eca\u9031\u3001\u4e3b\u8981\u306a\u65b0AI\u30e2\u30c7\u30eb\u3092\u767a\u8868\u4e88\u5b9a<\/h2>\n<ul>\n<li><strong>\u539f\u984c:<\/strong> SpaceXAI Will Reportedly Release a Major New AI Model This Week<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\">\u5c02\u9580\u30a2\u30ca\u30ea\u30b9\u30c8\u306e\u5206\u6790<\/h3>\n<div class=\"ai-summary-content\">\n<p><strong>SpaceXAI<\/strong>\uff08\u65e7<strong>xAI<\/strong>\uff09\u306f\u3001\u4eca\u9031\u4e2d\u306b\u65b0\u305f\u306a\u30d5\u30ed\u30f3\u30c6\u30a3\u30a2AI\u30e2\u30c7\u30eb\u3092\u767a\u8868\u3059\u308b\u3068\u5831\u3058\u3089\u308c\u3066\u3044\u307e\u3059\u3002\u3053\u306e\u30e2\u30c7\u30eb\u306f\u3001<strong>SpaceXAI<\/strong>\u3068\u3001\u73fe\u5728<strong>SpaceX<\/strong>\u304c600\u5104\u30c9\u30eb\u3067\u8cb7\u53ce\u3092\u9032\u3081\u3066\u3044\u308bAI\u30b9\u30bf\u30fc\u30c8\u30a2\u30c3\u30d7<strong>Cursor<\/strong>\u3068\u306e\u521d\u306e\u5171\u540c\u958b\u767a\u88fd\u54c1\u3068\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n<p><strong>Elon Musk<\/strong>\u6c0f\u306f\u5148\u6708\u3001\u672a\u516c\u958b\u306e\u6700\u65b0\u30e2\u30c7\u30eb<strong>Grok 4.5<\/strong>\u304c<strong>SpaceX<\/strong>\u3068<strong>Tesla<\/strong>\u3067\u30d7\u30e9\u30a4\u30d9\u30fc\u30c8\u30d9\u30fc\u30bf\u30c6\u30b9\u30c8\u4e2d\u3067\u3042\u308a\u3001\u305d\u306e\u6027\u80fd\u306f\u7279\u5b9a\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u306e<strong>Claude Opus<\/strong>\u306b\u5339\u6575\u3059\u308b\u3068<strong>X<\/strong>\u306b\u6295\u7a3f\u3057\u307e\u3057\u305f\u3002 <strong>The Information<\/strong>\u306e\u5831\u9053\u306b\u3088\u308b\u3068\u3001\u3053\u306e\u30e2\u30c7\u30eb\u306f\u793e\u5185\u3067<strong>Claude Opus 4.8<\/strong>\u304a\u3088\u3073<strong>OpenAI<\/strong>\u306e<strong>GPT 5.5<\/strong>\u3068\u6bd4\u8f03\u3055\u308c\u3066\u3044\u308b\u3068\u306e\u3053\u3068\u3067\u3059\u3002<\/p>\n<p>\ud83d\udc49 <strong><a href=\"https:\/\/gizmodo.com\/spacexai-will-reportedly-release-a-major-new-ai-model-this-week-2000782710\" target=\"_blank\" rel=\"noopener\">Gizmodo \u3067\u8a18\u4e8b\u5168\u6587\u3092\u8aad\u3080<\/a><\/strong><\/p>\n<\/div>\n<ul>\n<li><strong>\u8981\u70b9:<\/strong> SpaceXAI (xAI) is launching a new AI model, Grok 4.5, co-developed with Cursor, aiming to compete with top-tier models like Claude Opus and GPT 5.5.<\/li>\n<li><strong>\u8457\u8005:<\/strong> Mike Pearl<\/li>\n<\/ul>\n<blockquote class=\"wp-block-quote\"><p><span>English Summary:<\/span><\/p>\n<p><strong>SpaceXAI<\/strong> (formerly <strong>xAI<\/strong>) is reportedly launching a new frontier AI model as early as this week. This model is said to be the first joint production between <strong>SpaceXAI<\/strong> and <strong>Cursor<\/strong>, an AI startup that <strong>SpaceX<\/strong> is in the process of acquiring for $60 billion.<\/p>\n<p>Last month, <strong>Elon Musk<\/strong> posted on <strong>X<\/strong> that his company's latest unreleased model, <strong>Grok 4.5<\/strong>, based on a 1.5T V9 foundation model with <strong>Cursor<\/strong> data, was being beta tested privately at <strong>SpaceX<\/strong> and <strong>Tesla<\/strong>, with performance comparable to an unspecified version of <strong>Claude Opus<\/strong>. According to <strong>The Information<\/strong>, the model is being internally compared to <strong>Claude Opus 4.8<\/strong> and <strong>OpenAI's GPT 5.5<\/strong>.<\/p>\n<\/blockquote>\n<\/div>\n<div class=\"wp-block-group\" style=\"margin-top:40px;margin-bottom:40px\">\n<h2 class=\"wp-block-heading\">\u30d5\u30e9\u30f3\u30b9\u306e\u6ce8\u76ee\u30b9\u30bf\u30fc\u30c8\u30a2\u30c3\u30d7ZML\u3001\u591a\u6570\u306eAI\u30c1\u30c3\u30d7\u3067\u306e\u63a8\u8ad6\u3092\u9ad8\u901f\u5316\u3059\u308b\u7121\u6599\u88fd\u54c1\u3092\u30ea\u30ea\u30fc\u30b9<\/h2>\n<ul>\n<li><strong>\u539f\u984c:<\/strong> Hot French startup ZML releases free product to speed inference across lots of AI chips | TechCrunch<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\">\u5c02\u9580\u30a2\u30ca\u30ea\u30b9\u30c8\u306e\u5206\u6790<\/h3>\n<div class=\"ai-summary-content\">\n<p>\u30d5\u30e9\u30f3\u30b9\u306e\u6025\u6210\u9577\u4e2d\u306e\u30b9\u30bf\u30fc\u30c8\u30a2\u30c3\u30d7<strong>ZML<\/strong>\u304c\u3001\u591a\u6570\u306e<strong>AI\u30c1\u30c3\u30d7<\/strong>\u9593\u3067\u306e<strong>AI\u63a8\u8ad6<\/strong>\u3092\u9ad8\u901f\u5316\u3059\u308b\u305f\u3081\u306e\u7121\u6599\u88fd\u54c1\u3092\u30ea\u30ea\u30fc\u30b9\u3057\u307e\u3057\u305f\u3002\u3053\u306e\u88fd\u54c1\u306f\u3001\u5206\u6563\u578b\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u74b0\u5883\u306b\u304a\u3051\u308b<strong>AI\u30e2\u30c7\u30eb<\/strong>\u306e\u52b9\u7387\u7684\u306a\u5c55\u958b\u3068\u904b\u7528\u3092\u76ee\u7684\u3068\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n<p><strong>ZML<\/strong>\u306e\u3053\u306e\u53d6\u308a\u7d44\u307f\u306f\u3001<strong>AI<\/strong>\u306e\u904b\u7528\u30b3\u30b9\u30c8\u524a\u6e1b\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u5411\u4e0a\u306b\u8ca2\u732e\u3057\u3001\u3088\u308a\u5e83\u7bc4\u306a\u4f01\u696d\u304c\u9ad8\u5ea6\u306a<strong>AI<\/strong>\u6a5f\u80fd\u3092\u6d3b\u7528\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u3053\u3068\u3092\u76ee\u6307\u3057\u3066\u3044\u308b\u3068\u8003\u3048\u3089\u308c\u307e\u3059\u3002<\/p>\n<p>\ud83d\udc49 <strong><a href=\"https:\/\/techcrunch.com\/2026\/07\/08\/hot-french-startup-zml-releases-free-product-to-speed-inference-across-lots-of-ai-chips\/\" target=\"_blank\" rel=\"noopener\">TechCrunch \u3067\u8a18\u4e8b\u5168\u6587\u3092\u8aad\u3080<\/a><\/strong><\/p>\n<\/div>\n<ul>\n<li><strong>\u8981\u70b9:<\/strong> French startup ZML released a free product to significantly speed up AI inference across multiple AI chips, focusing on efficiency and distributed computing.<\/li>\n<li><strong>\u8457\u8005:<\/strong> Anna Heim<\/li>\n<\/ul>\n<blockquote class=\"wp-block-quote\"><p><span>English Summary:<\/span><\/p>\n<p>A rapidly emerging French startup named <strong>ZML<\/strong> has launched a free product designed to accelerate <strong>AI inference<\/strong> across numerous <strong>AI chips<\/strong>. This offering aims to enhance the efficient deployment and operation of <strong>AI models<\/strong> within distributed computing environments.<\/p>\n<p><strong>ZML's<\/strong> initiative is expected to contribute to reducing <strong>AI<\/strong> operational costs and improving performance, thereby enabling a broader range of companies to leverage advanced <strong>AI<\/strong> capabilities.<\/p>\n<\/blockquote>\n<\/div>\n<div class=\"wp-block-group\" style=\"margin-top:40px;margin-bottom:40px\">\n<h2 class=\"wp-block-heading\">Claude\u3068ChatGPT\u3001Microsoft\u306b\u3068\u3063\u3066\u3082\u9ad8\u4fa1\u306b\u306a\u308a\u3059\u304e\u3066\u3044\u308b<\/h2>\n<ul>\n<li><strong>\u539f\u984c:<\/strong> Claude and ChatGPT Are Getting Too Expensive, Even for Microsoft<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\">\u5c02\u9580\u30a2\u30ca\u30ea\u30b9\u30c8\u306e\u5206\u6790<\/h3>\n<div class=\"ai-summary-content\">\n<p><strong>Microsoft<\/strong>\u306f\u3001<strong>Claude<\/strong>\u3084<strong>ChatGPT<\/strong>\u306e\u3088\u3046\u306a\u6700\u5148\u7aef\u306e<strong>AI\u30e2\u30c7\u30eb<\/strong>\u306e\u5229\u7528\u30b3\u30b9\u30c8\u304c\u3001\u540c\u793e\u306b\u3068\u3063\u3066\u3082\u9ad8\u4fa1\u306b\u306a\u308a\u3059\u304e\u3066\u3044\u308b\u3068\u5831\u3058\u3089\u308c\u3066\u3044\u307e\u3059\u3002 <strong>AI\u30c8\u30fc\u30af\u30f3<\/strong>\u306e\u30b3\u30b9\u30c8\u304c\u9ad8\u9a30\u3057\u3066\u304a\u308a\u3001<strong>Microsoft<\/strong>\u306f\u8cbb\u7528\u524a\u6e1b\u306e\u305f\u3081\u306b\u81ea\u793e\u958b\u767a\u306e<strong>AI\u30e2\u30c7\u30eb<\/strong>\u306b\u5207\u308a\u66ff\u3048\u3066\u3044\u308b\u3068\u306e\u3053\u3068\u3067\u3059\u3002<\/p>\n<p><strong>Bloomberg<\/strong>\u306e\u5831\u9053\u306b\u3088\u308b\u3068\u3001<strong>Microsoft Excel<\/strong>\u304a\u3088\u3073<strong>Outlook<\/strong>\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u306b\u304a\u3051\u308b\u6bce\u9031\u6570\u4e07\u4ef6\u306e<strong>AI\u30d7\u30ed\u30f3\u30d7\u30c8<\/strong>\u304c\u3001\u73fe\u5728<strong>Microsoft<\/strong>\u72ec\u81ea\u306e<strong>MAI\u30e2\u30c7\u30eb<\/strong>\u3092\u4f7f\u7528\u3057\u3066\u5b8c\u4e86\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u30d7\u30ed\u30b0\u30e9\u30e0\u306f\u4ee5\u524d\u3001<strong>OpenAI<\/strong>\u3084<strong>Anthropic<\/strong>\u306e\u30e2\u30c7\u30eb\u306b\u5927\u304d\u304f\u4f9d\u5b58\u3057\u3066\u3044\u307e\u3057\u305f\u3002<\/p>\n<p><strong>Microsoft<\/strong>\u306f\u6700\u8fd1\u3001\u521d\u306e\u63a8\u8ad6\u30e2\u30c7\u30eb\u3067\u3042\u308b<strong>MAI-Thinking-1<\/strong>\u3092\u542b\u30807\u3064\u306e\u65b0\u3057\u3044\u793e\u5185\u30e2\u30c7\u30eb\u3092\u767a\u8868\u3057\u307e\u3057\u305f\u3002 \u3053\u306e\u30e2\u30c7\u30eb\u306f\u300c\u9ad8\u52b9\u7387\u3068\u9ad8\u6027\u80fd\u3001\u305d\u3057\u3066\u91cd\u8981\u306a\u3053\u3068\u306b\u4f4e\u30c8\u30fc\u30af\u30f3\u30b3\u30b9\u30c8\u300d\u306e\u305f\u3081\u306b\u69cb\u7bc9\u3055\u308c\u3001\u30d6\u30e9\u30a4\u30f3\u30c9\u30c6\u30b9\u30c8\u3067\u306f<strong>Anthropic<\/strong>\u306e<strong>Claude Opus 4.6<\/strong>\u306e\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u80fd\u529b\u306b\u5339\u6575\u3059\u308b\u3068\u3055\u308c\u3066\u3044\u307e\u3059\u3002 <strong>Microsoft AI<\/strong>\u306e<strong>CEO Mustafa Suleyman<\/strong>\u6c0f\u306f\u3001<strong>Anthropic<\/strong>\u3078\u306e\u652f\u6255\u3044\u3092\u524a\u6e1b\u3057\u3001\u6700\u7d42\u7684\u306b\u306f\u6392\u9664\u3059\u308b\u3053\u3068\u304c\u76ee\u6a19\u3060\u3068\u8ff0\u3079\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\ud83d\udc49 <strong><a href=\"https:\/\/gizmodo.com\/claude-and-chatgpt-are-getting-too-expensive-even-for-microsoft-2000782601\" target=\"_blank\" rel=\"noopener\">Gizmodo \u3067\u8a18\u4e8b\u5168\u6587\u3092\u8aad\u3080<\/a><\/strong><\/p>\n<\/div>\n<ul>\n<li><strong>\u8981\u70b9:<\/strong> Microsoft is shifting from expensive third-party AI models like Claude and ChatGPT to its own in-house MAI models, including the cost-efficient MAI-Thinking-1, to reduce high token costs.<\/li>\n<li><strong>\u8457\u8005:<\/strong> Bruce Gil<\/li>\n<\/ul>\n<blockquote class=\"wp-block-quote\"><p><span>English Summary:<\/span><\/p>\n<p><strong>Microsoft<\/strong> is reportedly finding the use of advanced <strong>AI models<\/strong> like <strong>Claude<\/strong> and <strong>ChatGPT<\/strong> to be too expensive, even for a company of its size. The cost of <strong>AI tokens<\/strong>, which measure computing work, is becoming so prohibitive that <strong>Microsoft<\/strong> is turning to its own models to save money.<\/p>\n<p>According to <strong>Bloomberg<\/strong>, tens of thousands of <strong>AI prompts<\/strong> each week in <strong>Microsoft's Excel<\/strong> and <strong>Outlook<\/strong> software are now being completed using <strong>Microsoft's<\/strong> own <strong>MAI models<\/strong>. These programs previously relied more heavily on models from <strong>OpenAI<\/strong> and <strong>Anthropic<\/strong>.<\/p>\n<p><strong>Microsoft<\/strong> recently unveiled seven new in-house models, including <strong>MAI-Thinking-1<\/strong>, its first reasoning model. This model was built for \"high efficiency and performance, but importantly, at a low-token cost,\" and in blind tests, it matched the coding abilities of <strong>Anthropic's<\/strong> popular <strong>Claude Opus 4.6<\/strong>. <strong>Microsoft AI CEO Mustafa Suleyman<\/strong> stated that their goal is to reduce and ultimately eliminate the significant costs paid to <strong>Anthropic<\/strong>.<\/p>\n<\/blockquote>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>Microsoft\u304cClaude\u3084ChatGPT\u306e\u9ad8\u30b3\u30b9\u30c8\u306b\u76f4\u9762\u3057\u3001\u81ea\u793eAI\u30e2\u30c7\u30eb\u3078\u306e\u79fb\u884c\u3092\u52a0\u901f\u3002SpaceXAI\u306fCursor\u3068\u306e\u5171\u540c\u958b\u767a\u306b\u3088\u308b\u65b0AI\u30e2\u30c7\u30ebGrok 4.5\u3092\u767a\u8868\u4e88\u5b9a\u3002\u30d5\u30e9\u30f3\u30b9\u306e\u30b9\u30bf\u30fc\u30c8\u30a2\u30c3\u30d7ZML\u306fAI\u63a8\u8ad6\u3092\u9ad8\u901f\u5316\u3059\u308b\u7121\u6599\u88fd\u54c1\u3092\u30ea\u30ea\u30fc\u30b9\u3002<\/p>\n","protected":false},"author":1,"featured_media":856,"comment_status":"","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"sns_share_botton_hide":"","vkExUnit_sns_title":"","_vk_print_noindex":"","sitemap_hide":"","_veu_custom_css":"","veu_display_promotion_alert":"","vkexunit_cta_each_option":"","_lightning_design_setting":[],"footnotes":""},"categories":[3],"tags":[8,461,16,76,502,503,501,26,62,15],"class_list":["post-1086","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-columns","tag-ai","tag-llm","tag-microsoft","tag-spacexai","tag-zml","tag-501","tag-26","tag-62"],"veu_head_title_object":{"title":"","add_site_title":""},"_links":{"self":[{"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/posts\/1086","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/comments?post=1086"}],"version-history":[{"count":0,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/posts\/1086\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/media\/856"}],"wp:attachment":[{"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/media?parent=1086"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/categories?post=1086"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/tags?post=1086"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}