Meta tags:
Headings (most frequently used words):
the, and, with, an, interview, about, of, ai, on, google, agentic, content, apple, china, parallel, founder, parag, agarwal, valuing, web, agents, inference, cook, in, tsmc, ceo, aws, sharp, vs, data, center, understanding, openai, ben, thompson, all, future, being, more, amazon, timing, mythos, muse, opportunity, compute, anthropic, microsoft, risk, world, models, spaghetti, discontent, opposition, fixing, problem, joanna, stern, living, sam, altman, matt, garman, bedrock, managed, articles, andrew, history, to, costs, preview, constructing, us, stability, trump, taiwan, comments, summit, takeaways, putin, note, goat, emergency, top, five, 2026, 21, veto, shift, durability, tim, impeccable, cost, 50, years, integration, over, bubbles, alignment, thin, is, software, survival, stratechery, updates, dithering, daring, fireball, john, gruber, asianometry, jon, yu, sinocism, bill, bishop, greatest, talk, tech, gpu, era, cerebras, implications, brief, training, neutrality, zero, one, doctrine, triumphs, marginal, meta, demand, supply, competitors, aggregates, real, threat, llm, paradigms, decreased, need, for, agency, enterprise, economic, imperatives, value, chain, department, war, north, korea, nuclear, weapons, complex, systems, who, entrust, ui, memory, crowd, out, beneficiaries, written, code, competition, work, iq, miss, token, foundries, brake, capex, plans, if, you, want, bubble, by, year, review,
Text of the page (most frequently used words):
the (1808), and (754), that (734), for (362), this (261), with (227), are (219), apple (186), more (173), not (169), they (169), was (169), but (167), will (165), its (141), you (140), have (136), from (131), which (129), about (128), has (127), company (112), what (107), their (106), one (106), all (103), there (89), compute (85), #models (82), can (82), anthropic (82), most (79), however (79), model (77), than (73), think (73), also (72), because (72), just (72), first (72), microsoft (72), much (68), when (67), amazon (67), openai (64), chips (64), other (64), would (64), software (62), companies (61), only (60), tsmc (60), make (60), like (60), were (59), been (58), may (57), going (56), memory (54), inference (54), business (53), time (53), over (53), use (53), google (51), actually (51), china (50), made (50), own (50), long (49), who (49), nvidia (49), being (48), need (48), how (48), both (48), those (48), then (48), agents (47), these (47), years (47), costs (47), new (46), year (45), means (45), demand (45), cost (44), cook (44), important (43), even (43), gpus (43), out (43), future (42), why (41), still (41), them (41), market (40), making (39), case (39), last (39), world (38), way (38), don (38), terms (37), isn (37), now (37), big (37), power (37), where (37), supply (36), our (36), point (36), customers (35), end (35), capacity (35), work (35), 2026 (34), chip (34), very (34), good (34), same (34), had (34), products (34), people (34), data (33), any (33), makes (32), well (32), agentic (31), some (31), listen (31), already (30), fact (30), could (30), enough (30), second (30), ceo (29), while (29), product (29), billion (29), many (29), used (29), into (28), want (28), another (28), should (28), aws (27), risk (27), take (27), right (27), tech (27), azure (27), better (27), stratechery (26), doesn (26), two (26), term (26), answer (26), every (26), training (26), problem (25), lot (25), least (25), put (25), particularly (25), increase (25), didn (25), question (25), run (25), best (25), iphone (25), large (25), get (24), might (24), needs (24), number (24), rather (24), given (24), app (24), code (24), pay (23), revenue (23), less (23), build (23), after (23), amodei (23), weapons (23), capabilities (23), claude (23), off (22), cloud (22), taiwan (21), value (21), marginal (21), growth (21), course (21), infrastructure (21), chatgpt (21), article (21), security (21), itself (21), hardware (21), consumer (21), interview (20), ever (20), does (20), said (20), built (20), three (20), far (20), services (20), device (20), third (20), reasoning (20), user (20), users (20), week (20), high (20), through (20), top (19), yes (19), biggest (19), hyperscalers (19), see (19), today (19), everyone (19), money (19), approach (19), your (19), clear (19), capex (19), yet (19), say (19), everything (19), call (19), post (19), humans (19), paradigm (19), content (18), note (18), back (18), potential (18), customer (18), few (18), thanks (18), impact (18), nuclear (18), here (18), enterprise (18), before (18), military (18), position (18), matter (18), agent (18), tokens (18), context (18), computer (18), jobs (18), frontier (18), integration (18), service (17), center (17), investment (17), industry (17), system (17), won (17), between (17), meta (17), using (17), copilot (17), opportunity (17), speed (17), systems (17), again (16), worth (16), competition (16), particular (16), down (16), building (16), above (16), without (16), earnings (16), able (16), including (16), such (16), advantage (16), technology (16), increasingly (16), comes (16), share (16), scale (16), seems (15), capital (15), massive (15), bubble (15), simply (15), bit (15), update (15), having (15), things (15), workloads (15), believe (15), networking (15), words (15), forward (15), single (15), major (15), certainly (15), chain (15), exactly (15), access (15), llms (15), intel (14), decision (14), entire (14), else (14), thing (14), ago (14), real (14), store (14), quarter (14), core (14), selling (14), intelligence (14), part (14), example (14), mean (14), shift (14), meanwhile (14), completely (14), internet (14), compelling (14), siri (14), local (14), labs (14), cerebras (14), likely (13), something (13), did (13), sure (13), wrote (13), invest (13), nearly (13), really (13), increased (13), look (13), chinese (13), theory (13), capability (13), his (13), watch (13), token (13), per (13), called (13), capable (13), history (13), coding (13), themselves (13), computing (13), mac (13), computers (13), useful (13), law (13), war (13), investments (13), mythos (13), view (12), plus (12), free (12), never (12), reality (12), face (12), era (12), performance (12), short (12), specifically (12), higher (12), ones (12), hand (12), silicon (12), source (12), explained (12), once (12), national (12), bloomberg (12), probably (12), log (12), podcast (12), later (12), argument (12), full (12), consider (12), devices (12), ultimately (12), government (12), department (12), meant (12), parallel (11), founder (11), web (11), hard (11), decade (11), meaningful (11), moreover (11), away (11), spending (11), significant (11), current (11), line (11), support (11), server (11), fast (11), sort (11), taking (11), needed (11), critical (11), came (11), deliver (11), apps (11), interface (11), place (11), results (11), reason (11), different (11), multiple (11), open (11), edge (11), available (11), reasons (11), bandwidth (11), windows (11), done (11), public (11), network (11), autonomous (11), under (11), step (11), agarwal (10), monday (10), ben (10), fully (10), offering (10), know (10), investing (10), truly (10), space (10), extension (10), productivity (10), bet (10), talking (10), extent (10), released (10), date (10), gemini (10), across (10), give (10), change (10), allow (10), economic (10), size (10), writing (10), larger (10), powerful (10), continues (10), soon (10), businesses (10), usage (10), relatively (10), team (10), lower (10), move (10), early (10), thick (10), thin (10), focused (10), price (10), centers (10), language (10), cases (10), deal (10), threat (10), during (10), around (10), true (10), llm (10), design (10), ibm (10), gpu (10), working (9), exist (9), beyond (9), basically (9), form (9), continue (9), whether (9), physical (9), got (9), executive (9), incredible (9), youtube (9), noted (9), feel (9), towards (9), doing (9), nature (9), together (9), fundamental (9), story (9), function (9), generated (9), distribution (9), leading (9), aggregation (9), competitive (9), rise (9), key (9), expensive (9), clients (9), become (9), consumers (9), smartphone (9), flow (9), matters (9), concept (9), debate (9), operating (9), united (9), surveillance (9), states (9), agency (9), profit (9), too (9), neo (9), cash (9), delivery (9), muse (9), progress (9), logistics (9), tuesday (8), thursday (8), parag (8), five (8), comments (8), thompson (8), game (8), sharp (8), search (8), rest (8), process (8), drive (8), moment (8), dollars (8), worse (8), lack (8), 2028 (8), dramatically (8), huge (8), significantly (8), plans (8), dependent (8), amount (8), generally (8), directly (8), interesting (8), implications (8), sell (8), integrated (8), write (8), applications (8), growing (8), human (8), etc (8), next (8), leverage (8), though (8), raises (8), base (8), paying (8), works (8), client (8), parts (8), focus (8), require (8), tools (8), chatbot (8), browser (8), android (8), international (8), iran (8), harness (8), found (8), inflection (8), announced (8), machines (8), became (8), valuing (7), trump (7), understanding (7), fixed (7), gets (7), hope (7), mobile (7), experience (7), potentially (7), challenge (7), aren (7), anything (7), spend (7), almost (7), looking (7), wasn (7), level (7), relative (7), asked (7), indeed (7), month (7), keep (7), favor (7), further (7), manufacturing (7), strategy (7), artificial (7), corp (7), north (7), korea (7), factor (7), serving (7), shows (7), applies (7), cpus (7), party (7), coming (7), must (7), let (7), via (7), written (7), changed (7), zero (7), advertising (7), since (7), day (7), according (7), million (7), tool (7), workflows (7), architecture (7), come (7), ultimate (7), set (7), tim (7), management (7), complex (7), possible (7), cheaper (7), platform (7), mostly (7), interest (7), either (7), control (7), launched (7), opus (7), initial (7), developers (7), getting (7), steve (7), managed (6), greatest (6), preview (6), friday (6), text (6), realize (6), harder (6), viable (6), competitors (6), understand (6), startup (6), returns (6), pure (6), play (6), foundry (6), created (6), critically (6), entailed (6), start (6), 2025 (6), expected (6), months (6), mistake (6), cfo (6), buy (6), obvious (6), attention (6), online (6), dominant (6), perhaps (6), advanced (6), chief (6), processors (6), ship (6), wall (6), margin (6), github (6), stack (6), innovation (6), balance (6), increasing (6), owning (6), seat (6), charge (6), area (6), driving (6), sense (6), within (6), faster (6), record (6), horizontal (6), saas (6), ecosystem (6), changing (6), against (6), absolutely (6), material (6), addressable (6), beginning (6), features (6), american (6), codex (6), recent (6), fits (6), whole (6), ended (6), electricity (6), entirely (6), task (6), anyone (6), arguably (6), electronics (6), huang (6), profits (6), smaller (6), limited (6), vertical (6), find (6), operate (6), vision (6), idea (6), points (6), defense (6), becomes (6), importance (6), imagine (6), serve (6), thousands (6), operations (6), loop (6), provide (6), ability (6), job (6), care (6), willing (6), modular (6), heavily (6), employees (6), video (6), required (6), weeks (6), traditional (6), introduced (6), difference (6), scaling (6), raw (6), timing (6), cache (6), altman (5), goat (5), emergency (5), constructing (5), stability (5), summit (5), takeaways (5), putin (5), opposition (5), updates (5), articles (5), older (5), sent (5), risks (5), reasonable (5), necessary (5), samsung (5), takes (5), prices (5), remember (5), facing (5), ask (5), worried (5), suddenly (5), managing (5), billions (5), hundreds (5), margins (5), trying (5), manage (5), wei (5), com (5), improve (5), spent (5), starting (5), represents (5), hype (5), haven (5), earlier (5), 2024 (5), providers (5), others (5), especially (5), pace (5), expect (5), side (5), situation (5), happened (5), moves (5), john (5), january (5), maker (5), similar (5), decisions (5), compete (5), street (5), resources (5), m365 (5), taken (5), benefit (5), thinking (5), seen (5), past (5), expanding (5), life (5), pressure (5), miss (5), valuable (5), fighting (5), delivering (5), application (5), suspect (5), publishers (5), instead (5), exclusive (5), great (5), commodity (5), running (5), majority (5), foundation (5), largest (5), march (5), launch (5), low (5), fraction (5), questions (5), somewhat (5), thought (5), addition (5), threatened (5), efforts (5), action (5), pcs (5), phones (5), started (5), commoditized (5), jensen (5), took (5), cutting (5), media (5), rival (5), corporate (5), combination (5), locally (5), utilization (5), weights (5), accomplish (5), profound (5), interaction (5), independent (5), output (5), developing (5), elected (5), president (5), address (5), concerns (5), person (5), based (5), deep (5), essential (5), basic (5), mass (5), cannot (5), choice (5), private (5), regardless (5), gives (5), uses (5), poised (5), rooted (5), dediu (5), enterprises (5), cowork (5), factory (5), always (5), precisely (5), macbook (5), internal (5), limit (5), create (5), chatbots (5), initiative (5), requires (5), generate (5), state (5), sora (5), inc (5), ipod (5), spark (5), distillation (5), widget (5), intensive (5), invested (5), gruber (5), spacex (5), existing (5), clouds (5), graphics (5), graviton (5), h100 (5), hbm (5), decode (5), april (4), living (4), talk (4), dithering (4), discontent (4), wednesday (4), spaghetti (4), explore (4), elon (4), issue (4), materialize (4), blown (4), whatever (4), faces (4), despite (4), decades (4), meet (4), quite (4), small (4), entails (4), fab (4), satisfy (4), extremely (4), depreciation (4), percentage (4), fabs (4), pricing (4), understandable (4), brake (4), secondly (4), planned (4), until (4), numbers (4), narrow (4), implication (4), shortage (4), driven (4), following (4), november (4), 2022 (4), told (4), solve (4), profitable (4), mentioned (4), amy (4), hood (4), adding (4), dependency (4), optimal (4), board (4), promise (4), analogy (4), developed (4), ending (4), brief (4), giving (4), dario (4), forum (4), davos (4), obviously (4), force (4), order (4), wanted (4), unique (4), weren (4), broader (4), issues (4), remains (4), option (4), nadella (4), investors (4), happy (4), defensive (4), gross (4), portfolio (4), constrained (4), grow (4), allocated (4), piece (4), solving (4), hiring (4), led (4), licensing (4), suite (4), makers (4), feels (4), 365 (4), delivers (4), latency (4), quality (4), happening (4), serious (4), justified (4), type (4), prospects (4), along (4), seemed (4), age (4), winners (4), quickly (4), original (4), return (4), focusing (4), suited (4), deterministic (4), role (4), ground (4), trajectory (4), unit (4), shares (4), stock (4), bigger (4), deepseek (4), bing (4), latest (4), provider (4), partnership (4), happens (4), path (4), happen (4), personal (4), reported (4), sony (4), components (4), related (4), self (4), fine (4), broadly (4), recently (4), contracts (4), versus (4), group (4), familiar (4), storage (4), superior (4), lab (4), rapidly (4), specific (4), underlying (4), largely (4), processing (4), argue (4), component (4), congress (4), safeguards (4), frame (4), weekend (4), consistently (4), says (4), easy (4), gains (4), sustainable (4), country (4), official (4), regard (4), development (4), effective (4), period (4), palantir (4), pentagon (4), decided (4), knowledge (4), adoption (4), values (4), dramatic (4), agreement (4), violation (4), wrong (4), operates (4), willingness (4), project (4), version (4), reducing (4), productive (4), changes (4), 100 (4), owns (4), maybe (4), hyperscaler (4), exponential (4), increases (4), fewer (4), host (4), implement (4), identified (4), wants (4), 3rd (4), iconic (4), upcoming (4), ive (4), pro (4), separate (4), subscriptions (4), dos (4), sold (4), developer (4), email (4), fortunate (4), laughlin (4), worry (4), 000 (4), digital (4), widgets (4), each (4), clearly (4), economics (4), virtue (4), doctrine (4), platforms (4), 2011 (4), stores (4), units (4), revolutionary (4), trainium (4), nitro (4), paas (4), prefill (4), wafer (4), serial (4), sam (3), matt (3), garman (3), bedrock (3), interviews (3), fixing (3), posts (3), beijing (3), success (3), portion (3), monopoly (3), rate (3), fails (3), wake (3), avoiding (3), shortages (3), exponentially (3), expense (3), definitely (3), sticks (3), challenges (3), exact (3), caught (3), modern (3), smartphones (3), amd (3), chance (3), anywhere (3), raising (3), overall (3), optimistic (3), benefits (3), fix (3), forgoing (3), competitor (3), offloaded (3), producing (3), bankrupt (3), dollar (3), revenues (3), easily (3), mind (3), disaster (3), notable (3), contribution (3), 2027 (3), little (3), 2029 (3), gap (3), several (3), admitted (3), accelerate (3), buildout (3), deals (3), saying (3), trend (3), committed (3), annual (3), 2023 (3), amongst (3), bottleneck (3), plan (3), aggressive (3), assumption (3), ways (3), mark (3), gcp (3), solutions (3), improved (3), remain (3), environment (3), expectations (3), exceeds (3), andy (3), argued (3), sale (3), destroyed (3), dead (3), positioned (3), priority (3), fair (3), neoclouds (3), oracle (3), proposition (3), appreciated (3), pre (3), lose (3), greater (3), expand (3), lifetime (3), allocating (3), scarce (3), acquiring (3), allocate (3), layers (3), guide (3), sales (3), accelerating (3), acceleration (3), downstream (3), predicted (3), functions (3), layer (3), individual (3), response (3), creating (3), bound (3), cover (3), cut (3), infinite (3), infinitely (3), valley (3), ipo (3), aggregators (3), articulation (3), newspaper (3), importantly (3), fundamentally (3), myself (3), newspapers (3), financial (3), correlated (3), tremendous (3), opposite (3), goes (3), bear (3), perfect (3), match (3), outputs (3), vast (3), longer (3), difficult (3), talked (3), engineers (3), admitting (3), laid (3), 2020 (3), skepticism (3), eventually (3), showed (3), slowed (3), members (3), factors (3), admit (3), functionality (3), disruption (3), gpt (3), help (3), gain (3), appears (3), turn (3), circle (3), concerned (3), hit (3), considered (3), button (3), thus (3), near (3), reached (3), improvements (3), crowd (3), cpu (3), impactful (3), passed (3), keynote (3), laptop (3), considering (3), drove (3), musk (3), runs (3), default (3), millions (3), dominance (3), dram (3), architectures (3), wave (3), request (3), result (3), chat (3), interfaces (3), premium (3), mainframe (3), send (3), unlike (3), interact (3), connectivity (3), felt (3), sun (3), although (3), java (3), phone (3), connected (3), room (3), 1980s (3), follow (3), daily (3), oversight (3), decide (3), consideration (3), responsible (3), controls (3), undergird (3), aspect (3), handle (3), americans (3), relationship (3), needless (3), democracy (3), status (3), independence (3), scenarios (3), putting (3), luckey (3), seeking (3), ongoing (3), agree (3), involved (3), apply (3), immediately (3), nation (3), democratic (3), leaving (3), categories (3), veto (3), hegseth (3), federal (3), designate (3), supplier (3), tilt (3), allowed (3), involvement (3), maduro (3), resist (3), stand (3), noting (3), answers (3), threats (3), stated (3), latter (3), labels (3), reliable (3), reliability (3), currently (3), outside (3), late (3), parties (3), entity (3), correct (3), commodities (3), bundle (3), differentiated (3), hasn (3), release (3), differentiation (3), earth (3), startups (3), sky (3), brilliant (3), hired (3), structures (3), drag (3), coordination (3), bottom (3), effectively (3), weaknesses (3), subscribers (3), processor (3), shipments (3), consumption (3), actively (3), direct (3), read (3), connection (3), generating (3), often (3), explain (3), verify (3), leaves (3), previously (3), tasks (3), continually (3), perspective (3), bubbles (3), former (3), angle (3), fifty (3), prediction (3), keeping (3), truth (3), win (3), monetize (3), assets (3), factories (3), class (3), fit (3), desktop (3), steps (3), ahead (3), news (3), jony (3), bonuses (3), develop (3), paid (3), deeply (3), music (3), lock (3), assistant (3), overhaul (3), markets (3), offerings (3), albeit (3), departments (3), multi (3), macintosh (3), 1984 (3), secure (3), six (3), primarily (3), missed (3), college (3), leveraging (3), memo (3), constraint (3), behind (3), widespread (3), primary (3), strategic (3), introduce (3), blog (3), widely (3), industrial (3), glasswing (3), commerce (3), calculations (3), servers (3), goods (3), stepping (3), gone (3), solution (3), massively (3), critique (3), priorities (3), schiller (3), introduction (3), expanded (3), program (3), university (3), distinct (3), trillion (3), satellite (3), drone (3), leo (3), explicitly (3), shipping (3), accelerator (3), emergence (3), times (3), semianalysis (3), iaas (3), database (3), freight (3), waiting (3), live (3), wse (3), colossus (3), andrew (3), twitter (3), learn (3), designed (2), privacy (2), joanna (2), stern (2), podcasts (2), popular (2), review (2), geographic (2), matches (2), capture (2), fabless (2), substantial (2), except (2), foregone (2), prove (2), surely (2), risky (2), hardly (2), centric (2), mythical (2), external (2), deployed (2), exists (2), defined (2), save (2), foundries (2), options (2), begging (2), declined (2), motivate (2), extra (2), losing (2), conservatism (2), inherently (2), flip (2), therefore (2), pressed (2), supported (2), projected (2), delayed (2), explains (2), notice (2), comparison (2), essentially (2), try (2), nervous (2), carefully (2), four (2), him (2), finally (2), believed (2), pushing (2), goals (2), 2021 (2), covid (2), basis (2), belief (2), cause (2), expenditures (2), smart (2), kind (2), kinds (2), pattern (2), zuckerberg (2), strong (2), tpus (2), cybersecurity (2), calls (2), exceeded (2), jassy (2), downside (2), realized (2), ought (2), semiconductor (2), equipment (2), administration (2), contrast (2), missiles (2), invasion (2), precariousness (2), believes (2), officer (2), blunder (2), h200 (2), editor (2), micklethwait (2), switzerland (2), crazy (2), reunification (2), figure (2), assume (2), prioritization (2), manufacturers (2), manufacture (2), contract (2), lines (2), raised (2), allocation (2), explicit (2), beat (2), 357 (2), misses (2), road (2), expands (2), moving (2), profile (2), dragon (2), remainder (2), sometimes (2), kpi (2), hopefully (2), helpful (2), opex (2), guidance (2), talented (2), teams (2), continued (2), replacement (2), her (2), prepared (2), identity (2), incorporating (2), compel (2), active (2), artifacts (2), organization (2), communications (2), boundary (2), powered (2), accurate (2), grounded (2), calling (2), satya (2), areas (2), players (2), necessarily (2), attacking (2), justifying (2), raise (2), analysis (2), tell (2), inputs (2), went (2), input (2), usable (2), broad (2), typical (2), overwhelming (2), abundance (2), fallen (2), reach (2), literally (2), glance (2), beneficiaries (2), list (2), standards (2), commitment (2), journey (2), nothing (2), competency (2), skeptical (2), perfectly (2), total (2), non (2), complaining (2), helped (2), feature (2), otherwise (2), arguing (2), paradigms (2), selloff (2), wiped (2), closed (2), 500 (2), contributed (2), worst (2), days (2), professional (2), briefly (2), subscription (2), recounting (2), placed (2), trick (2), information (2), immaterial (2), incredibly (2), inside (2), track (2), epoch (2), winner (2), intelligent (2), badly (2), llama (2), february (2), transition (2), aspects (2), prompts (2), seem (2), inevitable (2), crunch (2), foreseeable (2), alternative (2), sucks (2), consoles (2), playstation (2), telling (2), ps5 (2), wait (2), diminishing (2), sufficient (2), effect (2), pain (2), inevitably (2), ram (2), dynamic (2), purchases (2), show (2), ces (2), vera (2), rubin (2), announcement (2), energy (2), sector (2), oppo (2), targets (2), respond (2), requests (2), comment (2), console (2), generations (2), close (2), nintendo (2), card (2), leaders (2), crisis (2), laptops (2), sufficiently (2), breakthrough (2), lots (2), openclaw (2), orchestration (2), actual (2), practice (2), front (2), wherever (2), amounts (2), stronger (2), tread (2), golden (2), extreme (2), bustamante (2), conversation (2), licensed (2), looks (2), terminal (2), commands (2), understands (2), easier (2), field (2), click (2), computation (2), cheapest (2), victory (2), usually (2), innovations (2), downloaded (2), granted (2), central (2), push (2), keyboard (2), monitor (2), subject (2), limitations (2), nations (2), adversaries (2), tradeoffs (2), direction (2), pertinently (2), guns (2), align (2), laws (2), accountable (2), due (2), friction (2), resonates (2), consistent (2), safer (2), safety (2), incentives (2), unimaginable (2), wielded (2), minimum (2), fan (2), everybody (2), vocal (2), drivers (2), failure (2), notes (2), acknowledge (2), destroying (2), scenario (2), likelihood (2), equivalent (2), nobody (2), economy (2), leave (2), taiwanese (2), ukraine (2), conflict (2), looms (2), chains (2), resulted (2), discussion (2), pertinent (2), dominated (2), declared (2), pivot (2), asia (2), defend (2), lead (2), judgment (2), experiment (2), structure (2), outsourcing (2), bro (2), simple (2), please (2), asset (2), offensive (2), carrier (2), executives (2), alignment (2), insistence (2), rivals (2), sought (2), sympathetic (2), covered (2), offered (2), lets (2), deaths (2), middle (2), complicated (2), court (2), raid (2), semafor (2), removed (2), operation (2), senior (2), purpose (2), classified (2), secret (2), appearing (2), decides (2), nsa (2), provides (2), creative (2), legal (2), accede (2), lawful (2), remove (2), maintain (2), reserved (2), applied (2), act (2), puts (2), america (2), offer (2), proper (2), highly (2), trained (2), domestic (2), sources (2), safely (2), included (2), statement (2), stop (2), settings (2), journal (2), hold (2), onto (2), effectiveness (2), violating (2), serially (2), politics (2), opening (2), maintained (2), wind (2), couldn (2), goal (2), agnostic (2), possibility (2), organizational (2), permission (2), policies (2), seats (2), sustainably (2), betting (2), figuring (2), blow (2), limitation (2), unified (2), plausible (2), remotely (2), advice (2), understood (2), revolution (2), genius (2), moat (2), horace (2), imperatives (2), workforce (2), wonder (2), structurally (2), addressed (2), positive (2), machine (2), continuously (2), replacing (2), eliminating (2), apparatus (2), effort (2), demonstrated (2), bill (2), sit (2), alongside (2), convert (2), ends (2), pivoting (2), half (2), officially (2), certain (2), mainstream (2), tablets (2), utilized (2), observation (2), functional (2), fruit (2), industries (2), brownlee (2), scales (2), expenditure (2), usefulness (2), answering (2), couple (2), flaws (2), needing (2), flaw (2), directs (2), checks (2), fanfare (2), december (2), alternatives (2), mistakes (2), brilliance (2), doomed (2), known (2), september (2), tremendously (2), stream (2), hallucinate (2), count (2), proactively (2), head (2), opened (2), versions (2), transformer (2), impressive (2), october (2), sitting (2), gtc (2), conclusion (2), weird (2), vulnerability (2), competing (2), screen (2), materializes (2), ten (2), medium (2)
Text of the page (random words):
rticle entitled amazon s cloud crisis how aws will lose the future of computing and i found it very compelling first though some history much of which is covered in semianalysis article amazon not only invented cloud computing but also realized it would be a commodity market while most people in tech think about building sustainable differentiation that allows you to charge higher prices thus producing profit commodity markets work differently there sustainable profits come from having structurally cheaper costs amazon developed exactly that first through having the largest scale giving the company both buying power and also the most leverage on their development costs and second through genuine innovation aws built a specialized system called nitro built on their own chips that offloaded server management including network management storage management hypervisor management etc from the expensive intel and amd servers that the company sold access to this let amazon run that many more virtual machines on a single server significantly increasing utilization i e delivering a structural cost advantage amazon doubled down on their custom chip efforts with graviton their arm processors graviton chips particularly the first few generations were inferior to intel or amd chips but that didn t mean they were useless by that time aws had expanded from simply being an infrastructure as a service iaas provider to being a platform as a service paas provider as well iaas means you provide raw compute storage etc on which customers can run things like operating systems or databases paas means you provide that basic functionality as a service amazon relational database service rds for example is a fully managed database that customers can access via a set of apis without having to worry about actually managing the full database themselves worrying about scaling duplication etc this by extension means that customers don t need to know and don t need to care about the compute infrastructure that undergirds services like rds which has long been graviton paas lets amazon double dip in terms of profitability first aws could sell paas products at a higher margin than iaas products and second the company could leverage its own cheaper silicon to serve those products reducing their costs over time graviton has become more competitive in performance while still being cheaper giving amazon a lower cost compute instance to sell to end users but even without 3rd party take up the investment in building its own silicon has paid off over time training vs inference fast forward to ai and semianalysis concern was that all of these optimizations left aws ill prepared for ai one big problem was networking rather than implement the best networking from nvidia and or broadcom amazon is using its own nitro and elastic fabric adaptor efa networking this works well for many workloads plus it delivers a cost performance and security advantage there are business cultural and security reasons why amazon will not implement other networking the cultural one is important nitro and networking soc s generally have been amazon s biggest cost advantage for years it s ingrained into their dna even efa delivers on this too but they don t see how new workloads are evolving and that a new tier is needed due to the lack of foresight in their internal workload and infrastructure teams amazon is making a deliberate choice of not adopting that we believe will bite them in the future another was amazon s insistence on building its own chips which were not only inferior to the best nvidia chips in terms of performance but might also lead to them getting fewer nvidia chips going forward at least some other clouds will implement out of node nvlink that s where the discussion of prioritization now comes in ai gpus face tremendous shortages for at least a full year this is one of the most pivotal times for ai and it may mark the haves and the have nots nvidia is a complete monopoly right now why would nvidia prioritize amazon for these gpus when they know amazon will move to their in house chips as quickly as they can for as many compute workloads as they can why would nvidia ship tons of gpus to the cloud that is not using any of their networking thereby reducing their share of wallet instead nvidia prioritizes the me too clouds amazon does get meaningful volume but nowhere close to where demand is amazon s h100 gpu shipments relative to public cloud shipments is a significantly lower than their share of the public cloud those other clouds also can t satisfy demand but they get a bigger percentage of the gpus they ask nvidia for and as such firms looking for gpus for training or inference will move to those clouds nvidia is the kingmaker right now and they are capitalizing on it they have to spread the balance of power out to prevent compute share from clustering towards amazon these concerns were well founded in the 2023 time period when that article was written that was a time when ai thanks to chatgpt had hit the mainstream but the largest share of compute still went to training training required all of the things that amazon lacked particularly the ability to network large numbers of nvidia gpus together into one coherent system in such a system the most important capability was horizontal networking between chips so that you could update weights during training a step that needed to happen serially it was absolutely the case that cloud providers like microsoft or oracle or the neoclouds which implemented full nvidia solutions instead of the standalone hgx racks that aws favored were much better suited to training large language models that is still the case by the way what has changed is that training is no longer the biggest ai compute market inference is thanks not only to increased ai adoption but also because of fundamental changes in terms of how ai works from an update about nvidia the first inflection point was the emergence of llms call this the chatgpt moment in this first paradigm tokens were generated by gpus and presented as the answer to a question the second inflection point was the emergence of reasoning models call this the o1 moment in this paradigm there are a very large number of tokens that are generated to figure out the answer before the answer is actually generated this was an exponential increase in the addressable market for tokens the third inflection point was the emergence of functional agents call this the opus 4 5 moment in this paradigm those reasoning models are not triggered by humans asking a question but by an agent solving a problem this increases the market in two directions first humans can run multiple agents and secondly agents can leverage reasoning models multiple times to accomplish a task this isn t just an exponential increase in the addressable market for tokens it s two exponential increases squared both the shift to inference and the shift in the nature of inference have been positives for aws approach first while inference still requires significant memory the requirement is significantly less than that required for training it s actually viable to store a model s parameters in a single server you don t need to network together thousands of chips second while reasoning and agentic workloads require significantly more tokens and thus a massively larger kv cache the increase is actually so large that even the most optimized nvidia inference systems are being built with dedicated memory servers this sort of architecture is much more compatible with amazon s networking approach than the thousands of chips networked together approach is third agents are heavily cpu dependent which has two important implications first fully utilizing accelerators is a function of having sufficient general compute second achieving maximum utilization of heterogeneous compute means unbundling cpus and gpus and routing workloads between resources which is exactly the sort of disaggregated resource abstraction that amazon has been building with nitro the utilization point is an important one nvidia ceo jensen huang made his case for nvidia chips over custom asics at length at gtc 2025 huang s argument was that ai factories to use his term were ultimately constrained by power that meant that the most important metric for profitability was not the cost of chips but rather tokens per watt in other words if you can t increase watts it s worth spending more on chips to increase tokens on those watts there are however three reasons why this argument may not hold particularly for a company like amazon first if you have the money to buy that many nvidia chips you also have the money to spend on getting more power which is exactly what aws has been focused on this very much fits aws modus operandi which is to invest more upstream in this case in power with the goal of spending less downstream paying nvidia huge margins for their chips second in the long term electricity is more of a commodity than logic is that means it is a market where innovation and competition are more likely to break a bottleneck which is another way to say that investing in one s own silicon is the area most likely to deliver a return on investment third the nature of inference workloads particularly agentic ones is such that perfect accelerator utilization is going to be a much harder problem to solve than when it comes to training these points are moot however if you don t have your own logic chip that is at least competitive and here amazon s long term outlook is paying off amazon bought annapurna labs which makes their chips in 2015 and launched their first ai focused chip in 2019 no it wasn t very good but critically that was seven years ago now trainium 3 is decent and the trajectory is even better aws is positioned to have a sustainable cost advantage for inference going forward aws s neutrality moreover they are already replaying the graviton playbook trainium chips help undergird bedrock its ai platform which is to say that users are using trainium chips even if they didn t explicitly choose to do so aws ceo matt garman made this point explicitly in a stratechery interview i think just with gpus by the way you re going to interact with a lot of these accelerator chips through abstractions so the vast majority of customers don t interact with gpus either except through maybe like in their laptop or something like that for graphics but when you re talking to openai even if they re running on gpus you re not talking to the gpus if you re talking to claude you re through gpus or trainium or tpus you re not talking to any of those chips you re talking to the interface and the vast majority of inference out there is being done on one of a handful of models and so whether it s 5 10 20 100 it s not millions of people that are programming to those things directly and that s gonna be true going forward just because these systems are so complex they re very large if you re going to go train a model not that many people have enough money to go train a model not that many people have the expertise to actually manage it they re very complicated systems and the openai team is incredible in their ability to squeeze value out of a very large compute cluster but not that many people have the team that can do that independent of what the chip happens to be and so i think that that s going to be true for all accelerator chips honestly the frontier models are an important factor in this and that is an angle that i didn t see coming nvidia ceo jensen huang explained in a recent interview with dwarkesh patel why nvidia didn t invest in anthropic early on at the time i didn t deeply internalize how difficult it would be to build a foundation ai lab like openai and anthropic and the fact that they needed huge investments from the supplier themselves we just weren t in a position to make the multi billion dollar investment into anthropic so that they could use our compute but google and aws were they put in huge investments in the beginning so that anthropic in return used their compute we just weren t in a position to do that at the time i would say my mistake is i didn t deeply internalize that they really had no other options that a vc would never put in 5 10 billion of investment into an ai lab with the hopes of it turning out to be anthropic so that was my miss but even if i understood it i don t think we would ve been in a position to do that at the time but i m not going to make that same mistake again amazon had both the money and the chips to invest into anthropic precisely because they had built such a cash machine with aws in the first place that s the thing with big investments in infrastructure they take years to build but the benefit of that investment compounds over time anthropic meanwhile thanks to those investments from amazon and google can not only run across a variety of chips but for a long time was the only frontier model available on all of the leading clouds an important selling point for enterprises microsoft in the end needed to let go of azure s exclusive access to openai s api in part because that exclusivity was hurting the prospects of their mammoth stake in openai you can also make the case that amazon is the best choice for frontier model access in a world of limited compute microsoft s core business is software which is to say that the company faces massive pressure to invest in their own ai capabilities even at the cost of de prioritizing cloud customers that s exactly what happened at microsoft earlier this year when the company missed azure growth projections because they devoted more compute to their internal workloads it was an understandable decision cloud demand is eternal but the risk from ai for existing software businesses is existential this also applies to google the company s core business is also digital and while search has fended off the threat from chatbots that many expected the fundamental challenge is still one to be managed not extinguished amazon s core businesses meanwhile are very much rooted in the physical world selling and shipping physical goods and building data centers both are amenable to amazon devoting the majority of its chips to customers workloads amazon s future if this week marks the resolution of one of amazon s long bets you can see the outline of future resolutions in present day announcements one prominent example is amazon leo the company s satellite service that seems at first glance duplicative of spacex s starlink which has the advantage of already existing at scale remember amazon s formula however which ceo andy jassy stated explicitly with regards to leo on the company s most recent earnings call today if you ask what stops us from growing the business we have to get the constellation into space we have over 20 launches planned this year we have over 30 launches planned in 2027 but i think the business has a chance t...
|
