[{"data":1,"prerenderedAt":1461},["ShallowReactive",2],{"category-data-世界モデル":3},[4],{"id":5,"title":6,"body":7,"description":1439,"extension":1440,"meta":1441,"navigation":1456,"ogImage":1443,"path":1457,"seo":1458,"stem":1459,"__hash__":1460},"content/blogs/2-physical-ai-next-investment-theme.md","フィジカルAIが拓く次の投資フロンティア｜AMI Labsが示す世界モデルの可能性 | Physical AI: The Next Investment Frontier",{"type":8,"value":9,"toc":1389},"minimark",[10,715],[11,12,14,19,24,28,32,40,47,50,54,122,127,134,145,156,159,162,216,218,222,225,281,284,295,299,302,322,329,331,335,342,346,389,396,398,402,405,502,506,512,519,521,525,528,531,612,615,641,644,647,674,676,680,686,693,696,701,707,709],"lang-block",{"lang":13},"ja",[15,16,18],"h1",{"id":17},"フィジカルaiが拓く次の投資フロンティア","フィジカルAIが拓く次の投資フロンティア",[20,21,23],"h2",{"id":22},"ami-labsが示す世界モデルの可能性","——AMI Labsが示す「世界モデル」の可能性",[20,25,27],{"id":26},"はじめにllmの次に来るもの","はじめに：LLMの次に来るもの",[29,30,31],"p",{},"ChatGPTに代表される大規模言語モデル（LLM）が世界を席巻してから数年。生成AIへの投資ブームはひと段落し、投資家たちの視線は「その次」へと向かっています。",[29,33,34,35,39],{},"キーワードは ",[36,37,38],"strong",{},"フィジカルAI（Physical AI）"," ——AIが言語・画像だけでなく、現実の物理世界を理解し、その中で行動できるようになる技術領域です。",[29,41,42,43,46],{},"この潮流の象徴として世界が注目しているのが、チューリング賞受賞者・Yann LeCunが2026年初頭に創業した ",[36,44,45],{},"AMI Labs（Advanced Machine Intelligence Labs）"," です。",[48,49],"hr",{},[20,51,53],{"id":52},"ami-labsとは何か","AMI Labsとは何か",[55,56,57,70],"table",{},[58,59,60],"thead",{},[61,62,63,67],"tr",{},[64,65,66],"th",{},"項目",[64,68,69],{},"内容",[71,72,73,82,90,98,106,114],"tbody",{},[61,74,75,79],{},[76,77,78],"td",{},"正式名称",[76,80,81],{},"Advanced Machine Intelligence Labs（AMI Labs）",[61,83,84,87],{},[76,85,86],{},"創業",[76,88,89],{},"2026年（Yann LeCun × Alexandre Lebrun）",[61,91,92,95],{},[76,93,94],{},"本社",[76,96,97],{},"パリ（フランス）",[61,99,100,103],{},[76,101,102],{},"調達額",[76,104,105],{},"10.3億ドル（欧州史上最大のシードラウンド）",[61,107,108,111],{},[76,109,110],{},"バリュエーション",[76,112,113],{},"約35億ドル（pre-money）",[61,115,116,119],{},[76,117,118],{},"主要投資家",[76,120,121],{},"Cathay Innovation、Greycroft、HV Capital、Toyota Ventures、NVIDIA、Bezos Expeditions、Temasek ほか",[123,124,126],"h3",{"id":125},"技術の核心世界モデルworld-models","技術の核心：世界モデル（World Models）",[29,128,129,130,133],{},"AMI Labsが開発するのは、",[36,131,132],{},"世界の物理的な仕組みを学習するAI","——「世界モデル」です。",[29,135,136,137,140,141,144],{},"LeCunが2022年に提唱した ",[36,138,139],{},"JEPA（Joint Embedding Predictive Architecture）"," をベースとしており、LLMのようにテキストを生成するのではなく、",[36,142,143],{},"現実世界の抽象的な表現を学習し、行動の結果を予測・計画する"," ことが特徴です。",[146,147,152],"pre",{"className":148,"code":150,"language":151},[149],"language-text","LLMのアプローチ：　テキスト → テキスト生成\n世界モデルのアプローチ：センサーデータ → 物理世界の理解 → 行動計画\n","text",[153,154,150],"code",{"__ignoreMap":155},"",[29,157,158],{},"LeCunはかねてから「LLMだけでは汎用人工知能（AGI）には到達できない」と主張してきました。AMI Labsはその哲学を具現化した挑戦です。",[123,160,161],{"id":161},"主な応用分野",[55,163,164,174],{},[58,165,166],{},[61,167,168,171],{},[64,169,170],{},"分野",[64,172,173],{},"具体例",[71,175,176,184,192,200,208],{},[61,177,178,181],{},[76,179,180],{},"産業ロボティクス",[76,182,183],{},"工場での複雑な組み立て・検査作業",[61,185,186,189],{},[76,187,188],{},"自律走行",[76,190,191],{},"予測不能な道路環境への適応",[61,193,194,197],{},[76,195,196],{},"医療・ヘルスケア",[76,198,199],{},"手術支援ロボット、診断補助",[61,201,202,205],{},[76,203,204],{},"ウェアラブル",[76,206,207],{},"状況を理解して支援するデバイス",[61,209,210,213],{},[76,211,212],{},"宇宙・航空宇宙",[76,214,215],{},"遠隔環境での自律オペレーション",[48,217],{},[20,219,221],{"id":220},"なぜ今フィジカルaiなのか","なぜ今「フィジカルAI」なのか",[123,223,224],{"id":224},"市場規模の爆発的成長",[55,226,227,246],{},[58,228,229],{},[61,230,231,234,237,240,243],{},[64,232,233],{},"指標",[64,235,236],{},"2025年",[64,238,239],{},"2030年予測",[64,241,242],{},"2033年予測",[64,244,245],{},"CAGR",[71,247,248,265],{},[61,249,250,253,256,259,262],{},[76,251,252],{},"フィジカルAI市場",[76,254,255],{},"52億ドル",[76,257,258],{},"—",[76,260,261],{},"497億ドル",[76,263,264],{},"32.5%",[61,266,267,270,273,276,278],{},[76,268,269],{},"身体化AI（Embodied AI）",[76,271,272],{},"44億ドル",[76,274,275],{},"231億ドル",[76,277,258],{},[76,279,280],{},"39.0%",[123,282,283],{"id":283},"ヒューマノイドロボットの量産化",[29,285,286,287,290,291,294],{},"Goldman Sachsは2026年のヒューマノイドロボット世界出荷台数を ",[36,288,289],{},"5〜10万台"," と予測。製造コストも最終的には ",[36,292,293],{},"1台あたり1.5〜2万ドル"," まで低下すると見込まれており、大規模普及の条件が整いつつあります。",[123,296,298],{"id":297},"llmの次の壁","LLMの「次の壁」",[29,300,301],{},"LLMはテキストと画像の世界では驚異的な能力を発揮しましたが、以下の点で限界が露呈しています：",[303,304,305,312,317],"ul",{},[306,307,308,311],"li",{},[36,309,310],{},"物理世界の常識を理解できない","（コップを置けば落ちることを「知らない」）",[306,313,314],{},[36,315,316],{},"長期的な計画・行動が苦手",[306,318,319],{},[36,320,321],{},"ロボット操作などリアルタイム制御に不向き",[29,323,324,325,328],{},"世界モデルはこれらを克服し、",[36,326,327],{},"「考えて動く」AIを実現する鍵","として注目されています。",[48,330],{},[20,332,334],{"id":333},"scrum-venturesが注目するフィジカルai企業群","Scrum Venturesが注目するフィジカルAI企業群",[29,336,337,338,341],{},"サンフランシスコ・東京を拠点とするベンチャーキャピタル ",[36,339,340],{},"Scrum Ventures"," は、AIとロボティクスを主要投資テーマの一つに据え、フィジカルAI分野での先行投資を積極的に行っています。",[123,343,345],{"id":344},"注目ポートフォリオapptronik","注目ポートフォリオ：Apptronik",[55,347,348,356],{},[58,349,350],{},[61,351,352,354],{},[64,353,66],{},[64,355,69],{},[71,357,358,366,373,381],{},[61,359,360,363],{},[76,361,362],{},"企業名",[76,364,365],{},"Apptronik",[61,367,368,370],{},[76,369,170],{},[76,371,372],{},"ヒューマノイドロボット",[61,374,375,378],{},[76,376,377],{},"直近調達",[76,379,380],{},"3.5億ドル（B Capital・Googleが主導）",[61,382,383,386],{},[76,384,385],{},"特徴",[76,387,388],{},"産業用途向け二足歩行ロボット「Apollo」を開発。Nasaとの共同研究実績も持つ",[29,390,391,392,395],{},"Scrum Venturesはこのような、",[36,393,394],{},"現実の物理環境で動作するAIロボット企業","への投資を通じて、フィジカルAI時代のインフラ構築を支援しています。",[48,397],{},[20,399,401],{"id":400},"フィジカルai主要プレイヤー比較","フィジカルAI主要プレイヤー比較",[29,403,404],{},"AMI Labsと同様に、世界モデル・ロボット基盤モデルを開発する主要企業を比較します。",[55,406,407,422],{},[58,408,409],{},[61,410,411,414,417,419],{},[64,412,413],{},"企業",[64,415,416],{},"調達総額",[64,418,110],{},[64,420,421],{},"技術的特徴",[71,423,424,440,456,471,487],{},[61,425,426,431,434,437],{},[76,427,428],{},[36,429,430],{},"AMI Labs",[76,432,433],{},"10.3億ドル",[76,435,436],{},"約35億ドル",[76,438,439],{},"世界モデル（JEPA）、LeCun主導",[61,441,442,447,450,453],{},[76,443,444],{},[36,445,446],{},"Physical Intelligence（π）",[76,448,449],{},"11億ドル",[76,451,452],{},"約56億ドル",[76,454,455],{},"ロボット基盤モデル（π0）、家庭・産業用",[61,457,458,462,465,468],{},[76,459,460],{},[36,461,365],{},[76,463,464],{},"3.5億ドル以上",[76,466,467],{},"非公開",[76,469,470],{},"ヒューマノイドロボット「Apollo」、NASA連携",[61,472,473,478,481,484],{},[76,474,475],{},[36,476,477],{},"Figure AI",[76,479,480],{},"6.75億ドル",[76,482,483],{},"約26億ドル",[76,485,486],{},"OpenAIとの提携、Figure 01/02",[61,488,489,494,497,499],{},[76,490,491],{},[36,492,493],{},"1X Technologies",[76,495,496],{},"1億ドル以上",[76,498,467],{},[76,500,501],{},"OpenAIバックアップの人型ロボット",[123,503,505],{"id":504},"ami-labs-vs-physical-intelligence-の違い","AMI Labs vs. Physical Intelligence の違い",[146,507,510],{"className":508,"code":509,"language":151},[149],"AMI Labs：\n  └── 「世界を理解するAI」が先。ロボットはその応用の一つ\n  └── 基礎研究主導。JEPA で物理法則の抽象表現を学習\n\nPhysical Intelligence (π)：\n  └── 「ロボットを動かすAI」に特化\n  └── π0 モデルをロボットメーカーに提供（OpenAI的ポジション）\n",[153,511,509],{"__ignoreMap":155},[29,513,514,515,518],{},"両社は補完的な関係にあり、",[36,516,517],{},"世界モデル（AMI）× ロボット操作モデル（π）"," の組み合わせが、将来的な汎用ロボットを実現する可能性があります。",[48,520],{},[20,522,524],{"id":523},"投資家視点フィジカルaiをどう捉えるか","投資家視点：フィジカルAIをどう捉えるか",[123,526,527],{"id":527},"バリューチェーンで考える",[29,529,530],{},"フィジカルAI産業は以下のレイヤーに分解できます：",[55,532,533,545],{},[58,534,535],{},[61,536,537,540,542],{},[64,538,539],{},"レイヤー",[64,541,69],{},[64,543,544],{},"代表企業・銘柄",[71,546,547,560,573,586,599],{},[61,548,549,554,557],{},[76,550,551],{},[36,552,553],{},"基盤モデル",[76,555,556],{},"世界モデル・ロボット基盤AI",[76,558,559],{},"AMI Labs、Physical Intelligence",[61,561,562,567,570],{},[76,563,564],{},[36,565,566],{},"ハードウェア",[76,568,569],{},"センサー、チップ、アクチュエーター",[76,571,572],{},"NVIDIA（NVDA）、Mobileye",[61,574,575,580,583],{},[76,576,577],{},[36,578,579],{},"ロボット本体",[76,581,582],{},"ヒューマノイド・産業ロボット",[76,584,585],{},"Apptronik、Figure AI、Boston Dynamics",[61,587,588,593,596],{},[76,589,590],{},[36,591,592],{},"プラットフォーム",[76,594,595],{},"データ収集・シミュレーション基盤",[76,597,598],{},"Foxglove（40M調達）、Dyna Robotics",[61,600,601,606,609],{},[76,602,603],{},[36,604,605],{},"エンドユーザー",[76,607,608],{},"製造・物流・医療への導入",[76,610,611],{},"Amazon、Tesla、Toyota",[123,613,614],{"id":614},"リスク要因",[303,616,617,623,629,635],{},[306,618,619,622],{},[36,620,621],{},"技術的不確実性","：世界モデルの実用化は研究段階のものが多い",[306,624,625,628],{},[36,626,627],{},"ハードウェアのボトルネック","：ロボット本体のコスト・耐久性問題",[306,630,631,634],{},[36,632,633],{},"規制・安全基準","：医療・公道での自律機械への規制が整備途上",[306,636,637,640],{},[36,638,639],{},"長い開発サイクル","：ソフトウェアAIと異なり、実証・量産に時間がかかる",[123,642,643],{"id":643},"個人投資家の現実的アプローチ",[29,645,646],{},"フィジカルAIのスタートアップへの直接投資は上場前企業が多く困難です。現実的なアプローチ：",[648,649,650,656,662,668],"ol",{},[306,651,652,655],{},[36,653,654],{},"NVIDIA（NVDA）","：フィジカルAIの「ピッケルと鍬」。GPUに加え、Isaacsim（ロボットシミュレーター）でエコシステムを構築",[306,657,658,661],{},[36,659,660],{},"ロボティクス関連ETF","：ROBO、ARKQ など",[306,663,664,667],{},[36,665,666],{},"トヨタ・ホンダ等の自動車メーカー株","：自律走行・工場自動化への先行投資",[306,669,670,673],{},[36,671,672],{},"間接的受益企業","：センサーメーカー、半導体設計会社",[48,675],{},[20,677,679],{"id":678},"まとめフィジカルaiは10年に一度の転換点","まとめ：フィジカルAIは「10年に一度」の転換点",[29,681,682,683,46],{},"LLMが「情報処理革命」だとすれば、フィジカルAIは ",[36,684,685],{},"「物理世界との相互作用革命」",[29,687,688,689,692],{},"AMI Labsの10億ドル調達は、単なる一スタートアップの成功ではなく、",[36,690,691],{},"AI産業の重心がデジタル空間から物理空間へと移行する歴史的な転換点のシグナル","です。",[29,694,695],{},"Scrum Venturesのような先見性のある投資家が、この波に早期からベットしている事実は示唆に富みます。",[29,697,698],{},[36,699,700],{},"ZYL0 の視点：",[702,703,704],"blockquote",{},[29,705,706],{},"フィジカルAIは「SFの世界」ではなく、今まさに資本が流れ込んでいる「現実の投資テーマ」です。\n基盤モデルとハードウェアのコスト低下が交差する2026〜2028年が、参入タイミングの鍵を握ると見ています。",[48,708],{},[29,710,711],{},[712,713,714],"em",{},"免責事項：本記事は情報提供を目的としており、投資アドバイスを構成するものではありません。すべての投資判断はご自身の判断と資格を有するファイナンシャルアドバイザーへの相談に基づいて行ってください。",[11,716,718,722,726,730,733,740,747,749,753,815,819,826,836,842,845,849,903,905,909,913,965,969,980,984,987,1005,1011,1013,1017,1022,1026,1068,1075,1077,1081,1173,1177,1183,1190,1192,1196,1200,1203,1284,1288,1314,1318,1321,1347,1349,1353,1363,1369,1372,1377,1382,1384],{"lang":717},"en",[15,719,721],{"id":720},"physical-ai-the-next-investment-frontier","Physical AI: The Next Investment Frontier",[20,723,725],{"id":724},"what-ami-labs-reveals-about-the-world-model-era","——What AMI Labs Reveals About the World Model Era",[20,727,729],{"id":728},"introduction-what-comes-after-llms","Introduction: What Comes After LLMs?",[29,731,732],{},"Large language models (LLMs) like ChatGPT have dominated the technology landscape for the past several years. Now, as the initial generative AI investment frenzy settles, the smart money is asking a new question: what comes next?",[29,734,735,736,739],{},"The answer is increasingly clear: ",[36,737,738],{},"Physical AI"," — the technology domain where AI moves beyond language and images to understand and act within the real, physical world.",[29,741,742,743,746],{},"The defining signal of this shift came in early 2026, when Turing Award winner Yann LeCun launched ",[36,744,745],{},"AMI Labs (Advanced Machine Intelligence Labs)",", securing the largest seed round in European history and placing a very large bet on a fundamentally different kind of AI.",[48,748],{},[20,750,752],{"id":751},"what-is-ami-labs","What Is AMI Labs?",[55,754,755,765],{},[58,756,757],{},[61,758,759,762],{},[64,760,761],{},"Item",[64,763,764],{},"Details",[71,766,767,775,783,791,799,807],{},[61,768,769,772],{},[76,770,771],{},"Full Name",[76,773,774],{},"Advanced Machine Intelligence Labs (AMI Labs)",[61,776,777,780],{},[76,778,779],{},"Founded",[76,781,782],{},"2026 (Yann LeCun × Alexandre Lebrun)",[61,784,785,788],{},[76,786,787],{},"Headquarters",[76,789,790],{},"Paris, France",[61,792,793,796],{},[76,794,795],{},"Funding",[76,797,798],{},"$1.03B (Europe's largest seed round ever)",[61,800,801,804],{},[76,802,803],{},"Valuation",[76,805,806],{},"~$3.5B (pre-money)",[61,808,809,812],{},[76,810,811],{},"Key Investors",[76,813,814],{},"Cathay Innovation, Greycroft, HV Capital, Toyota Ventures, NVIDIA, Bezos Expeditions, Temasek, and others",[123,816,818],{"id":817},"the-core-technology-world-models","The Core Technology: World Models",[29,820,821,822,825],{},"AMI Labs is building AI that learns how the physical world works — so-called ",[36,823,824],{},"world models",".",[29,827,828,829,832,833,825],{},"The approach is grounded in ",[36,830,831],{},"JEPA (Joint Embedding Predictive Architecture)",", a framework LeCun proposed in 2022 as an alternative to the LLM paradigm. Rather than generating text, JEPA-based systems learn abstract representations of real-world sensory data and use them to ",[36,834,835],{},"predict consequences and plan sequences of actions",[146,837,840],{"className":838,"code":839,"language":151},[149],"LLM approach:    Text → Generate more text\nWorld Model:     Sensor data → Understand physical world → Plan actions\n",[153,841,839],{"__ignoreMap":155},[29,843,844],{},"LeCun has long argued that LLMs alone cannot reach artificial general intelligence. AMI Labs is his answer to that gap.",[123,846,848],{"id":847},"target-applications","Target Applications",[55,850,851,861],{},[58,852,853],{},[61,854,855,858],{},[64,856,857],{},"Domain",[64,859,860],{},"Examples",[71,862,863,871,879,887,895],{},[61,864,865,868],{},[76,866,867],{},"Industrial robotics",[76,869,870],{},"Complex assembly, quality inspection in factories",[61,872,873,876],{},[76,874,875],{},"Autonomous vehicles",[76,877,878],{},"Adapting to unpredictable road environments",[61,880,881,884],{},[76,882,883],{},"Healthcare",[76,885,886],{},"Surgical assist robots, diagnostic support",[61,888,889,892],{},[76,890,891],{},"Wearables",[76,893,894],{},"Context-aware assistive devices",[61,896,897,900],{},[76,898,899],{},"Aerospace",[76,901,902],{},"Autonomous operations in remote environments",[48,904],{},[20,906,908],{"id":907},"why-physical-ai-why-now","Why Physical AI, Why Now?",[123,910,912],{"id":911},"market-size-explosive-growth-ahead","Market Size: Explosive Growth Ahead",[55,914,915,933],{},[58,916,917],{},[61,918,919,922,925,928,931],{},[64,920,921],{},"Metric",[64,923,924],{},"2025",[64,926,927],{},"2030 Forecast",[64,929,930],{},"2033 Forecast",[64,932,245],{},[71,934,935,950],{},[61,936,937,940,943,945,948],{},[76,938,939],{},"Physical AI Market",[76,941,942],{},"$5.2B",[76,944,258],{},[76,946,947],{},"$49.7B",[76,949,264],{},[61,951,952,955,958,961,963],{},[76,953,954],{},"Embodied AI Market",[76,956,957],{},"$4.4B",[76,959,960],{},"$23.1B",[76,962,258],{},[76,964,280],{},[123,966,968],{"id":967},"humanoid-robots-are-entering-mass-production","Humanoid Robots Are Entering Mass Production",[29,970,971,972,975,976,979],{},"Goldman Sachs forecasts global humanoid robot shipments of ",[36,973,974],{},"50,000–100,000 units in 2026",", with per-unit costs eventually declining to ",[36,977,978],{},"$15,000–$20,000",". The conditions for mass adoption are falling into place.",[123,981,983],{"id":982},"the-wall-llms-cannot-climb","The Wall LLMs Cannot Climb",[29,985,986],{},"LLMs have been extraordinary at language and image tasks, but they have clear structural limitations:",[303,988,989,995,1000],{},[306,990,991,994],{},[36,992,993],{},"No physical common sense"," (they don't \"know\" that a cup left on a ledge will fall)",[306,996,997],{},[36,998,999],{},"Weak at long-horizon planning and physical action",[306,1001,1002],{},[36,1003,1004],{},"Ill-suited for real-time robotic control",[29,1006,1007,1008,825],{},"World models are designed to overcome these limitations — enabling AI that can ",[36,1009,1010],{},"reason about and act within the physical world",[48,1012],{},[20,1014,1016],{"id":1015},"physical-ai-companies-on-scrum-ventures-radar","Physical AI Companies on Scrum Ventures' Radar",[29,1018,1019,1021],{},[36,1020,340],{},", a seed-stage VC with offices in San Francisco and Tokyo, has identified AI and robotics as a core investment thesis. The firm is among those actively deploying capital into the Physical AI space.",[123,1023,1025],{"id":1024},"portfolio-spotlight-apptronik","Portfolio Spotlight: Apptronik",[55,1027,1028,1036],{},[58,1029,1030],{},[61,1031,1032,1034],{},[64,1033,761],{},[64,1035,764],{},[71,1037,1038,1045,1052,1060],{},[61,1039,1040,1043],{},[76,1041,1042],{},"Company",[76,1044,365],{},[61,1046,1047,1049],{},[76,1048,857],{},[76,1050,1051],{},"Humanoid robotics",[61,1053,1054,1057],{},[76,1055,1056],{},"Recent Funding",[76,1058,1059],{},"$350M (led by B Capital and Google)",[61,1061,1062,1065],{},[76,1063,1064],{},"Technology",[76,1066,1067],{},"Apollo biped robot for industrial environments; NASA collaboration track record",[29,1069,1070,1071,1074],{},"Scrum Ventures' involvement in companies like Apptronik reflects a thesis that ",[36,1072,1073],{},"the AI-native robotic stack"," — hardware + software + world models — will define the next computing platform.",[48,1076],{},[20,1078,1080],{"id":1079},"competitive-landscape-key-physical-ai-players","Competitive Landscape: Key Physical AI Players",[55,1082,1083,1096],{},[58,1084,1085],{},[61,1086,1087,1089,1092,1094],{},[64,1088,1042],{},[64,1090,1091],{},"Total Raised",[64,1093,803],{},[64,1095,1064],{},[71,1097,1098,1113,1129,1144,1159],{},[61,1099,1100,1104,1107,1110],{},[76,1101,1102],{},[36,1103,430],{},[76,1105,1106],{},"$1.03B",[76,1108,1109],{},"~$3.5B",[76,1111,1112],{},"World models (JEPA), Yann LeCun",[61,1114,1115,1120,1123,1126],{},[76,1116,1117],{},[36,1118,1119],{},"Physical Intelligence (π)",[76,1121,1122],{},"$1.1B",[76,1124,1125],{},"~$5.6B",[76,1127,1128],{},"Robot foundation models (π0)",[61,1130,1131,1135,1138,1141],{},[76,1132,1133],{},[36,1134,365],{},[76,1136,1137],{},"$350M+",[76,1139,1140],{},"Undisclosed",[76,1142,1143],{},"Humanoid robot \"Apollo\", NASA partner",[61,1145,1146,1150,1153,1156],{},[76,1147,1148],{},[36,1149,477],{},[76,1151,1152],{},"$675M",[76,1154,1155],{},"~$2.6B",[76,1157,1158],{},"OpenAI partnership, Figure 01/02",[61,1160,1161,1165,1168,1170],{},[76,1162,1163],{},[36,1164,493],{},[76,1166,1167],{},"$100M+",[76,1169,1140],{},[76,1171,1172],{},"OpenAI-backed humanoid robots",[123,1174,1176],{"id":1175},"ami-labs-vs-physical-intelligence-a-key-distinction","AMI Labs vs. Physical Intelligence: A Key Distinction",[146,1178,1181],{"className":1179,"code":1180,"language":151},[149],"AMI Labs:\n  └── \"Understanding the world\" is the goal; robots are one application\n  └── Research-led. JEPA learns abstract physical representations\n\nPhysical Intelligence (π):\n  └── Focused specifically on \"making robots act\"\n  └── Provides π0 model to robotics manufacturers (OpenAI-style positioning)\n",[153,1182,1180],{"__ignoreMap":155},[29,1184,1185,1186,1189],{},"These two companies are more complementary than competitive. The combination of ",[36,1187,1188],{},"world models (AMI) × robot action models (π)"," may be the path to truly general-purpose robots.",[48,1191],{},[20,1193,1195],{"id":1194},"investment-framework-how-to-think-about-physical-ai","Investment Framework: How to Think About Physical AI",[123,1197,1199],{"id":1198},"the-value-chain","The Value Chain",[29,1201,1202],{},"Physical AI breaks down into distinct investment layers:",[55,1204,1205,1217],{},[58,1206,1207],{},[61,1208,1209,1212,1215],{},[64,1210,1211],{},"Layer",[64,1213,1214],{},"Description",[64,1216,860],{},[71,1218,1219,1232,1245,1258,1271],{},[61,1220,1221,1226,1229],{},[76,1222,1223],{},[36,1224,1225],{},"Foundation Models",[76,1227,1228],{},"World models, robot action AI",[76,1230,1231],{},"AMI Labs, Physical Intelligence",[61,1233,1234,1239,1242],{},[76,1235,1236],{},[36,1237,1238],{},"Hardware",[76,1240,1241],{},"Sensors, chips, actuators",[76,1243,1244],{},"NVIDIA (NVDA), Mobileye",[61,1246,1247,1252,1255],{},[76,1248,1249],{},[36,1250,1251],{},"Robots",[76,1253,1254],{},"Humanoid, industrial machines",[76,1256,1257],{},"Apptronik, Figure AI, Boston Dynamics",[61,1259,1260,1265,1268],{},[76,1261,1262],{},[36,1263,1264],{},"Platforms",[76,1266,1267],{},"Data infra, simulation tools",[76,1269,1270],{},"Foxglove ($40M), Dyna Robotics",[61,1272,1273,1278,1281],{},[76,1274,1275],{},[36,1276,1277],{},"End Users",[76,1279,1280],{},"Manufacturing, logistics, healthcare",[76,1282,1283],{},"Amazon, Tesla, Toyota",[123,1285,1287],{"id":1286},"key-risk-factors","Key Risk Factors",[303,1289,1290,1296,1302,1308],{},[306,1291,1292,1295],{},[36,1293,1294],{},"Technical uncertainty",": World models are still largely in research stages",[306,1297,1298,1301],{},[36,1299,1300],{},"Hardware bottlenecks",": Robot body cost, durability, reliability",[306,1303,1304,1307],{},[36,1305,1306],{},"Regulatory landscape",": Autonomous machines in healthcare and public spaces face evolving rules",[306,1309,1310,1313],{},[36,1311,1312],{},"Long development cycles",": Unlike software AI, physical systems require real-world validation and manufacturing scale-up",[123,1315,1317],{"id":1316},"realistic-strategies-for-individual-investors","Realistic Strategies for Individual Investors",[29,1319,1320],{},"Direct investment in Physical AI startups is largely inaccessible pre-IPO. Practical approaches:",[648,1322,1323,1329,1335,1341],{},[306,1324,1325,1328],{},[36,1326,1327],{},"NVIDIA (NVDA)",": The \"picks and shovels\" of Physical AI. Beyond GPUs, Isaac Sim (robot simulator) makes NVIDIA the ecosystem hub",[306,1330,1331,1334],{},[36,1332,1333],{},"Robotics ETFs",": ROBO, ARKQ",[306,1336,1337,1340],{},[36,1338,1339],{},"Automotive OEMs",": Toyota, Honda — investing heavily in autonomous manufacturing and mobility",[306,1342,1343,1346],{},[36,1344,1345],{},"Indirect beneficiaries",": Sensor manufacturers, chip design companies",[48,1348],{},[20,1350,1352],{"id":1351},"conclusion-physical-ai-is-a-once-in-a-decade-shift","Conclusion: Physical AI Is a Once-in-a-Decade Shift",[29,1354,1355,1356,1359,1360,825],{},"If LLMs represented a revolution in ",[36,1357,1358],{},"information processing",", Physical AI represents a revolution in ",[36,1361,1362],{},"interaction with the physical world",[29,1364,1365,1366,825],{},"AMI Labs' $1B raise is not just one startup's success story — it is a ",[36,1367,1368],{},"signal that the center of gravity of the AI industry is shifting from digital space to physical space",[29,1370,1371],{},"The fact that forward-thinking investors like Scrum Ventures are placing bets on this thesis early is worth noting.",[29,1373,1374],{},[36,1375,1376],{},"ZYL0's View:",[702,1378,1379],{},[29,1380,1381],{},"Physical AI is not science fiction — it is an investment theme where real capital is flowing right now.\nThe convergence of declining hardware costs and maturing foundation models makes 2026–2028 the critical window to watch.",[48,1383],{},[29,1385,1386],{},[712,1387,1388],{},"Disclaimer: This article is for informational purposes only and does not constitute investment advice. All investment decisions should be made based on your own judgment and in consultation with a qualified financial advisor.",{"title":155,"searchDepth":1390,"depth":1390,"links":1391},2,[1392,1393,1394,1399,1404,1407,1410,1415,1416,1417,1418,1422,1427,1430,1433,1438],{"id":22,"depth":1390,"text":23},{"id":26,"depth":1390,"text":27},{"id":52,"depth":1390,"text":53,"children":1395},[1396,1398],{"id":125,"depth":1397,"text":126},3,{"id":161,"depth":1397,"text":161},{"id":220,"depth":1390,"text":221,"children":1400},[1401,1402,1403],{"id":224,"depth":1397,"text":224},{"id":283,"depth":1397,"text":283},{"id":297,"depth":1397,"text":298},{"id":333,"depth":1390,"text":334,"children":1405},[1406],{"id":344,"depth":1397,"text":345},{"id":400,"depth":1390,"text":401,"children":1408},[1409],{"id":504,"depth":1397,"text":505},{"id":523,"depth":1390,"text":524,"children":1411},[1412,1413,1414],{"id":527,"depth":1397,"text":527},{"id":614,"depth":1397,"text":614},{"id":643,"depth":1397,"text":643},{"id":678,"depth":1390,"text":679},{"id":724,"depth":1390,"text":725},{"id":728,"depth":1390,"text":729},{"id":751,"depth":1390,"text":752,"children":1419},[1420,1421],{"id":817,"depth":1397,"text":818},{"id":847,"depth":1397,"text":848},{"id":907,"depth":1390,"text":908,"children":1423},[1424,1425,1426],{"id":911,"depth":1397,"text":912},{"id":967,"depth":1397,"text":968},{"id":982,"depth":1397,"text":983},{"id":1015,"depth":1390,"text":1016,"children":1428},[1429],{"id":1024,"depth":1397,"text":1025},{"id":1079,"depth":1390,"text":1080,"children":1431},[1432],{"id":1175,"depth":1397,"text":1176},{"id":1194,"depth":1390,"text":1195,"children":1434},[1435,1436,1437],{"id":1198,"depth":1397,"text":1199},{"id":1286,"depth":1397,"text":1287},{"id":1316,"depth":1397,"text":1317},{"id":1351,"depth":1390,"text":1352},"Yann LeCunが創業したAMI Labsを軸に、フィジカルAI・世界モデルが次の巨大投資テーマとなる理由を解説。Scrum Venturesが注目する具体的な企業群と投資戦略を日英バイリンガルで紹介。","md",{"date":1442,"image":1443,"alt":1444,"tags":1445,"tagsEn":1451,"published":1456},"21st Mar 2026","/blogs-img/blog-physical-ai.png","フィジカルAIと世界モデル：次世代投資テーマの全貌",[1446,1447,1448,1449,1450],"フィジカルAI","AI投資","ロボティクス","世界モデル","スタートアップ",[738,1452,1453,1454,1455],"AI Investment","Robotics","World Models","Startups",true,"/blogs/2-physical-ai-next-investment-theme",{"title":6,"description":1439},"blogs/2-physical-ai-next-investment-theme","pWGz4PbzyzsFQIpJeKfMvMPVb977qCF-ovg98P1t_aU",1775926673410]