Read the latest
Tuesday, August 25, 2026HotTea archive editionVerified 9:08 AM PDT

5 minutes. Facts before narrative.

OpenAI posted Jalapeno's first inference numbers

OpenAI says Jalapeno used less power and answered faster than comparison systems. If those numbers hold in production, OpenAI pays less for each answer and gets more bargaining room with suppliers. The AI bill now runs through chips, power, funding, and export rules.

Start the briefingSee how we verify →

Published daily by 3:00 AM Pacific. No forced optimism. No manufactured panic.

Listen to today’s briefing

OpenAI posted Jalapeno's first inference numbers

The sourced HotTea edition, condensed into a chaptered morning podcast with verified audio and a full transcript.

Follow on Apple PodcastsFollow on SpotifyPodcast home →
01

The lead

Models, products, infrastructure, energy, companies, and marketsOpenAI wants each answer to cost less.

OpenAI posted Jalapeno's first inference results

OpenAI said Jalapeno is its first custom inference chip. The company said the chip delivered 1.5 to 1.9 times more AI work per watt. OpenAI also claimed latency was 1.7 to 3.6 times lower than comparison systems across three public models.

What happened

OpenAI posted the first measured Jalapeno results on August 25. The company said it built the chip for inference, especially interactive agent work. OpenAI said it plans to add Jalapeno to its compute infrastructure by the end of 2026.

Why it matters

OpenAI pays to serve each AI answer. If its numbers hold outside the release, Jalapeno could cut latency and power use. The chip could also give OpenAI more room in supplier negotiations.

What to watch

Watch independent InferenceX results and real deployment volume. Watch the latency users feel. Jalapeno changes OpenAI's economics only if it takes enough production work off the cloud bill or the Nvidia bill.

The caveat

OpenAI supplied the benchmark claims. The company compared Jalapeno with commercial systems on public workloads. Those results do not prove that Jalapeno changes OpenAI's production costs.

OpenAI: Jalapeno's first results show industry-leading speed and efficiency in AI inference OpenAI: The full stack behind abundant intelligence
Read this story on its own →

Worth knowing

The rest of the morning

Facts, pressure point, next evidence.

02
AI geopolitics, export controls, war, security, companies, and marketsChip controls can fail inside the sales chain.

Taiwan charged nine people over alleged AI-server smuggling

Taiwanese prosecutors charged nine people on Monday over alleged illegal exports of high-end AI servers to mainland China. The group included one Nvidia employee and two former Super Micro employees. AP reported that 74 servers reached China, and Taiwanese authorities stopped 56 more.

Pressure point Governments can write chip rules. Sales still run through people. Staff approvals, site checks, paperwork, and resellers can become the weak points.

Watch Watch criminal penalties in Taiwan. Watch whether U.S. agencies cite the case in future licensing rules. New compliance controls from Nvidia or Super Micro would show that the case changed company behavior.

Associated Press ↗Financial Times ↗
Read article →
03
Infrastructure, energy, macroeconomics, politics, and regulationAI demand is turning into gas demand.

AI data centers are bringing U.S. gas power back

Global Energy Monitor said U.S. gas power capacity in development rose from 252 GW in January to 378 GW by mid-2026. The Guardian reported that the United States is now building twice as much gas-fired capacity as China. Wired reported that gas projects built just for data centers grew from 97 GW at the end of 2025 to more than 189 GW.

Pressure point Many projects may never get built. Grid delays could still push companies toward private gas plants first. Ratepayers and local communities may face the fuel, emissions, and permitting fights later.

Watch Watch which projects get financing. Watch whether utilities separate data-center costs in public filings. Project-level power and emissions terms from tech companies would show who pays.

Global Energy Monitor ↗The Guardian ↗Wired ↗
Read article →
04
Politics, regulation, science, society, companies, and marketsA child-safety lawsuit could become product policy.

Meta's child-safety trial could force product changes

The Guardian reported that California and 28 other states are suing Meta in Oakland. The states allege Facebook and Instagram were designed to addict children and hide the harm. Meta denies the allegations and says the states are chasing an outlandish payout.

Pressure point This is not an AI model case. AI products still sit near the same legal line. If courts force changes to social platforms, states may try similar claims against chatbots, recommender systems, and child-safety design.

Watch Watch whether the court orders product changes. Watch whether internal Meta documents become public. State attorneys general could move beyond Meta if they use the same theory against AI companions.

The Guardian ↗Associated Press ↗
Read article →
05
Companies, markets, AI geopolitics, export controls, and macroeconomicsAI spending is hitting shareholders now.

Alibaba sold new shares to fund AI infrastructure

Alibaba said it priced an HK$80 billion placement of new shares in Hong Kong. The company plans to use all net proceeds for full-stack AI capabilities, including infrastructure. MarketWatch reported that the sale was the largest primary follow-on sale by a Chinese company.

Pressure point Alibaba is asking public investors to fund more chips, models, and infrastructure before the profit path is clear. U.S. hyperscalers face the same question. Alibaba also brings China exposure and shareholder dilution.

Watch Watch whether the placement closes on August 26. Watch whether Alibaba breaks out returns from AI infrastructure. Qwen adoption would be stronger proof if it offsets cloud margin pressure.

Alibaba Group ↗MarketWatch ↗
Read article →
06
Companies, markets, macroeconomics, infrastructure, and energyAI stocks still need proof.

AI stocks fell before Nvidia's earnings

Investor's Business Daily reported that Nvidia and several memory stocks sold off before Nvidia's Wednesday earnings report. July personal-consumption inflation data was also due later in the week. Investor's Business Daily reported separately that Alibaba's U.S.-listed shares fell after the company priced its AI share sale.

Pressure point Companies are spending on AI, and the market already knows it. Investors now need revenue, margins, and inflation conditions that can carry that spending at current valuations.

Watch Watch Nvidia's data-center revenue and memory pricing. Watch capital-spending guidance from large buyers. July inflation could change interest-rate conditions for AI stocks that depend on future growth.

Investor's Business Daily ↗Investor's Business Daily ↗
Read article →

The whole AI power map

AI is no longer a tech beat.

HotTea follows where AI moves power, money, labor, security, and state capacity—not only where a new model scores higher.

01

Politics & regulation

Elections, procurement, courts, surveillance, lobbying, and state power.

02

Economics & labor

Productivity, wages, employment, capital spending, concentration, and who captures the gains.

03

War & security

Autonomy, cyber operations, intelligence, targeting, export controls, and escalation risk.

04

AI geopolitics

Chips, energy, alliances, sovereign capability, supply chains, and strategic competition.

05

Markets & companies

Funding, revenue, margins, model economics, enterprise adoption, and infrastructure bets.

06

Science & society

Medicine, education, climate, culture, research, rights, and measurable public outcomes.

Infrastructure, capital, and regulation

AI now needs chips, power, rules, and money

Five pressure points told the same story today: inference hardware, export controls, gas power, courts, and capital markets.

1

OpenAI's chip claims show why model labs want more control over inference costs before agents become everyday products.

2

Taiwan's charges and the U.S. gas buildout show that AI capacity now depends on enforcement paperwork and power choices, not only chips.

3

Alibaba's placement and the AI-stock drop show that public markets are still funding the buildout. Investors want proof sooner.

The watchlist

Signals that could change the read

WatchlistIndependent Jalapeno benchmarks, plus proof that OpenAI is using Jalapeno in production.Tracking
WatchlistTaiwan or U.S. rule changes after the alleged B300 server shipments.Tracking
WatchlistData-center gas projects that move from announced capacity to funded construction.Tracking
WatchlistNvidia earnings details that separate AI demand from inventory, memory pricing, and power limits.Tracking

How HotTea works

No optimism quota. No negativity quota. Just the sourced read.

Every reported item links to its source. Company claims remain company claims. High-risk stories require stronger corroboration. Material caveats, conflicts, and unknowns stay in the story. HotTea’s interpretation is visibly separated so readers can disagree without losing the facts.

Edition validated · 6 stories · 13 unique sources

Audit today’s sources →

Tomorrow’s signal, before tomorrow’s noise

Open HotTea. Know what changed.

A new verified edition every morning. If the evidence or release gate fails, the last verified briefing stays live.

Back to today’s top ↑