

Google · v8 · depuis Angekündigt am 22. April 2026 (Google Cloud Next); allgemeine Verfügbarkeit laut Google "später in diesem Jahr" (Status: · 7× · vu le 30 juin 2026
"TPU 8" is Google's eighth-generation Tensor Processing Unit family, split for the first time into two specialized chips instead of one unified design: TPU 8t for large-scale model training and TPU 8i for low-latency inference and agentic workloads. Both were unveiled on April 22, 2026 at Google Cloud Next, run on Google's own Arm-based Axion CPU hosts, and use fourth-generation liquid cooling. TPU 8t scales to 9,600 chips per "superpod" delivering 121 FP4 ExaFLOPs, while TPU 8i delivers 11.6 ExaFLOPs FP8 in pods of 1,152 chips. Both chips are officially listed as not yet generally available ("Coming soon"), with wider availability expected later in 2026, though some tech outlets report a 2027 GA target; official pricing has not been published by Google.
Fonctionnalités
| Manufacturing Process (nm) | TSMC N3 process family (3 nm) |
| License | Proprietary Google Cloud hardware, exclusive to Google Cloud customers (not sold to third parties) |
| Platform | Google Cloud AI Hypercomputer, hosted on Google Axion Arm-based CPUs |
| Price | Not published by Google; third-party analysts estimate TPU 8i at roughly $5-7 per chip-hour at GA |
| Compute Performance (FLOPS/TOPS) | TPU 8t: 121 FP4 ExaFLOPs per superpod (9,600 chips); TPU 8i: 11.6 ExaFLOPs FP8 per pod (1,152 chips) |
| Release Date | Announced April 22, 2026 at Google Cloud Next |
| Memory | TPU 8t: 216 GB HBM3e per chip (6,528 GB/s), 2 PB shared memory per superpod; TPU 8i: 288 GB HBM3e + 384 MB on-chip SRAM per chip |
| Availability | Not yet generally available ('Coming soon'); Google targets later in 2026 |
Plus de produits dans cette catégorie: Matériel d'entraînement IA
Preuves (7)
Company Analysis: Google
HOLD is the most data-consistent stance given (1) strong operating momentum evidenced by Q2 2026 results, (2) supportive sell-side consensus/targets, but (3) unusually high policy/remedy execution risk in both the EU (DMA fine and specified compliance measures) and the US (ad tech behavioral relief), plus (4) elevated uncertainty around AI-driven capex and Cloud margin durability highlighted by Visible Alpha commentary. (sec.gov)
Summary
Alphabet (Google) is a diversified digital platform company whose core cash engine remains advertising (Google Search, YouTube, Network), complemented by subscription/device revenue (e.g., YouTube subscriptions, Google One, hardware) and a rapidly scaling infrastructure/software business in Google Cloud. Its core competencies are (1) global-scale consumer distribution (Search, Android, Chrome, Maps, YouTube), (2) data/ML infrastructure and model development, and (3) monetization via performance advertising and cloud services. Market position remains top-tier in global search and large in digital video via YouTube, with competitive advantages rooted in default distribution, product integration, and advertiser ROI. However, the moat is increasingly constrained by regulation and remedies: in the EU, the Commission issued a €890m DMA fine (Search self-preferencing and Play steering) and is specifying compliance measures for Google Search data access under DMA Article 6(11). (digital-strategy.ec.europa.eu) In the US, the DOJ announced “substantial relief” in the ad tech monopolization case, including behavioral remedies affecting product integration and interoperability. (justice.gov) These items are directly relevant to medium-term monetization mechanics and take-rate stability. On fundamentals, Alphabet reported strong Q2 2026 results (released July 22, 2026), underscoring resilient ad demand and continued profitability. (sec.gov) While the company does not provide traditional quarterly revenue guidance in the press release, investor focus in recent months has shifted to (a) Cloud operating margin trajectory and (b) AI-related capex intensity and returns. S&P Global/Visible Alpha noted that consensus revenue expectations rose into Q2, but operating income/EPS expectations were tempered by lower Cloud margin assumptions; it also highlighted rising capex expectations tied to AI infrastructure and a debate about whether Cloud margins have peaked. (spglobal.com) Valuation and consensus positioning: Street consensus remains constructive, with S&P Global-polled analysts showing a “Strong Buy” consensus and an average 12‑month price target around $429 (USD) as of mid‑September 2026; forward P/E metrics cited by aggregators cluster in the mid‑20s depending on the earnings basis and horizon. (stockanalysis.com) For an investor benchmarking in EUR, the equity is still USD‑denominated (GOOG/GOOGL), so EUR returns will be meaningfully influenced by EURUSD moves. Short- to medium-term outlook is best framed as: (1) operating performance supported by ad resilience and Cloud scale economics, offset by (2) regulatory/remedy execution risk (EU DMA and US ad tech remedies) and (3) capex/ROI scrutiny as AI infrastructure spending rises. Net: fundamentals are strong, but the risk-adjusted path depends on how quickly Alphabet can adapt product design and monetization to new compliance constraints while sustaining Cloud profitability improvements. (justice.gov)
Key Takeaways
- Alphabet’s earnings power remains anchored in Search/YouTube advertising, with Cloud increasingly important to incremental margin and narrative support. (sec.gov)
- Regulatory pressure is no longer abstract: the EU issued a €890m DMA fine and specified compliance measures for Google Search data access; these can affect ranking presentation, traffic allocation, and monetization. (digital-strategy.ec.europa.eu)
- US ad tech remedies are progressing, with DOJ describing significant behavioral relief; implementation details can influence ad tech take rates and integration advantages. (justice.gov)
- Consensus remains positive (Strong Buy; average target ~$429 USD), but valuation is sensitive to Cloud margin assumptions and capex/ROI expectations. (stockanalysis.com)
- For EUR-based investors, currency translation is a first-order variable because the stock is USD-priced; EUR performance can diverge from underlying USD equity performance even if fundamentals track expectations. (stockanalysis.com)
Action Ideas
12–24 month accumulation thesis: Use Alphabet as a large-cap compounder exposure where (i) core ads remain resilient and (ii) Cloud profitability can continue to scale, while the market debates capex intensity. The data-driven support is that consensus remains constructive (S&P Global-polled “Strong Buy” and ~$429 average target as of Sep 2026), and Visible Alpha commentary indicates revenue expectations have been rising even as margin expectations are debated—creating potential for upside if Cloud margins stabilize above reduced expectations. For EUR investors, consider staging entries and explicitly managing EURUSD exposure (unhedged vs hedged) because FX can dominate near-term EUR returns.
Horizon: 18 mo.
6–12 month risk-managed hold: Maintain exposure but avoid adding until there is clearer evidence on (a) Cloud margin trajectory after recent consensus downticks and (b) the operational impact/timing of EU DMA and US ad tech remedies. This is appropriate if you already have a position and want to reduce decision risk while monitoring the next earnings cycle and compliance milestones. The rationale is that consensus is supportive, but the dispersion of outcomes is elevated due to regulatory implementation and capex/ROI debates.
Horizon: 9 mo.
3–6 month de-risking thesis (tactical): Reduce exposure if your portfolio cannot tolerate regulatory headline risk and potential multiple compression tied to capex/ROI concerns. The data basis is that both EU and US actions have moved into concrete remedy/fine stages in 2026, which can create event-driven volatility independent of near-term revenue delivery. This is not a call on business deterioration; it is a risk-budget decision when regulatory outcomes are the dominant uncertainty driver.
Horizon: 4 mo.
Contrarian Insights
- • Regulatory actions may be a nearer-term earnings variable than AI competition: the EU DMA fine plus specified Search data-access measures and the US ad tech remedies are concrete, dated, and implementation-driven—potentially affecting monetization mechanics sooner than many AI-disruption narratives imply. (digital-strategy.ec.europa.eu)
- • Cloud margin expectations may be more fragile than consensus positioning suggests: Visible Alpha commentary highlights that Cloud margin estimates have already been revised down and show a very wide range, implying that small changes in cost assumptions (AI infra, depreciation, staffing) can materially change consolidated earnings leverage even if revenue remains strong. (spglobal.com)
Sources (7)
- https://www.sec.gov/Archives/edgar/data/1652044/000165204426000066/googexhibit991q22026.htm
- https://www.justice.gov/opa/pr/department-justice-again-wins-substantial-relief-against-google
- https://digital-strategy.ec.europa.eu/en/news/commission-fines-google-eu890-million-breaches-digital-markets-act
- https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX%3A52026DMA100209
- https://www.spglobal.com/market-intelligence/en/news-insights/research/2026/07/alphabet-earnings-preview-q2-2026
- https://stockanalysis.com/stocks/googl/forecast/
- https://www.trefis.com/data/companies/GOOG