サプライヤー

[片山幹健（Tomotake Katayama) @tomotake94]

ソフトバンクが、スイスのロボット開発スタートアップ、グラビス・ロボティクスの買収を検討している。企業価値は5億ドル超（約820億円）。
新設するAI・ロボティクス企業「Roze」の傘下で保有する方針。

グラビスは2022年、スイス連邦工科大学チューリヒ校のスピンオフとして設立。土木工事用の建設機械を自律運転させるハードウェアとソフトウェアを開発しています。

ソフトバンクのロボティクス資産は倉庫（バークシャー・グレイ）、産業用アーム（ABB・アジャイル）、基盤モデル（スキルドAI）がありますが、今回のグラビスによって建設・土木という新たな産業ドメインが加わります。

https://t.co/KkD4tbtIG8

[サプライヤー @dzlso]

@Kansai_MNP 知りたいです

[AGIラボ @ctgptlb]

【速報】中国発のAIがFable 5/GPT 5.6 Sol超え

朝からビッグニュースが飛び込んできました。なんと中国企業Moonshotが開発したAIモデル、Kimi K3の性能が、米国の最先端AIを抑えてAI Agent Evaluationsなどの分野で1位を獲得。そして、オープンウェイトは2026年7月27日までに公開予定』とのこと

[平川 凌 @g7z5e]

コメ兵ホールディングス、モノ資産の管理アプリ「cashari（カシャリ）」を運営するガレージバンクの全株式を16.8億円で取得
https://t.co/ZXDS35DRf0

ガレージバンク 2025年12月期
売上高10億円、営業損失2億円

2025年6月末の（潜在株含む）調達後評価額は、31億円

[平川 凌 @g7z5e]

ヒューリックが2025年4月から展開している「こどもでぱーと」
https://t.co/pCz67t4VRE

「子どもの教育」と「親の負担軽減」を両立するために1つの建物にあらゆるコンテンツを集約している。多子世帯にとってはめちゃくちゃ嬉しい気がする

現在は2棟のみの運営だが、すでに2028年までで5棟の開業を予定しており、29年には20棟へ増やす目標を掲げている

[Melvin @MelvinInvests]

You missed the semiconductor trade because you were afraid, you missed the memory trade because you were afraid and now you are going to miss the next rally for the exact same reason (Save this).

Look at what just happened this week alone.

ASML beat Q2 2026 estimates and raised its full year revenue outlook to €43-45 billion, up from prior guidance of €36-40 billion, the second time this year they have raised guidance.

Their CEO said on the call that "ongoing AI-related investments and continued progress in AI technologies are driving demand for advanced logic and memory chips" and that their customers are "accelerating their capacity expansion plans."

Memory revenue alone is expected to grow 75% this year.

TSMC raised 2026 capex guidance to $52-56 billion, at least 25% above 2025 levels and guided for nearly 30% revenue growth.

Meanwhile, Meta is doubling its data center deployments in 2027.

OpenAI's GPT-5.6 compute literally cannot keep up with demand, Codex went from 5 million to 9 million users in a single month, and Sam Altman confirmed agentic product usage rose 2.5x in a single week.

Anthropic is approaching profitability meaning even the companies spending the most on inference compute are starting to turn the corner on monetization, which means the demand cycle extends further.

This is the pattern and it repeats every single time.

In 2023, people missed Nvidia because they said valuations were too high and AI was overhyped meanwhile Nvidia hit 5T in market cap.

In early 2025, people missed Micron at $80 because they said the memory cycle was peaking while Micron crossed $1 trillion in market cap.

The pattern is always the same and the data is always clear.

The fear is always louder than the data and the people who let the fear win are always the ones watching from the sidelines when the rally resumes.

The AI thesis is completely intact, hyperscalers are still guiding to $1.4 trillion in capex by 2028.

The foundational infrastructure, chips, memory, data centers, power is being built out faster than any previous technology cycle in history.

The applications consuming that infrastructure, from autonomous coding agents to physical AI to agentic workflows, are growing faster than even the most optimistic projections.

What is keeping the market down right now is also some Iran war noise and rate hike fears.

Neither of those things changes the fact that Meta is doubling its data centers, ASML is raising guidance for the second time this year, TSMC is spending $56 billion in capex, Micron's HBM is sold out through 2027, and OpenAI's infrastructure literally cannot keep up with how fast people are using it.

This is a temporary dip within a stronger upcycle and the people who act on the data instead of the fear are the ones who will look back at this week as one of the better buying opportunities of the year.

I am a buyer at these levels and make sure to follow me @MelvinInvests for more market insights.

[Mark Kretschmann @mark_k]

What @Kimi_Moonshot pulled off here with Kimi K3 is nothing short of an earthquake in the AI industry. Kimi K3 just took the #1 spot on the Frontend Code Arena with a score of 1,679, beating Claude Fable 5, GPT-5.6 Sol, GLM-5.2, Grok-4.5, and every Claude Opus model.

And this is not a narrow win. Kimi K3 is 48 points ahead of second place and 61 points ahead of GPT-5.6 Sol. Moonshot AI is no longer merely catching up with the frontier labs. It is setting the pace!

[Chetaslua @chetaslua]

🚨 Kimi k3 is on par with Fable-5 and GPT-5.6 Sol

we are so back now

[Chubby♨️ @kimmonismus]

Kimi K3 may be the DeepSeek 2.0 moment. The benchmark results are out, and they are outstanding. And they proof one thing very clearly.

I believe this is the DeepSeek 2.0 moment. Does that sound like an exaggeration? At first glance, perhaps. The benchmarks still need to prove themselves in real-world use. But one thing already seems clear: Kimi K3 is ahead of Opus 4.8 and, broadly speaking, just behind GPT-5.6 and Fable 5.

Opus 4.8 was released at the end of May, specifically on May 26, roughly a month and a half ago. Kimi K3 generally outperforms Opus 4.8. That effectively disproves the claim that Chinese open source models consistently trail US closed source models by six to eight months.

Of course, most users will continue to choose the very best model available, and I do not expect millions of people to cancel their ChatGPT subscriptions overnight. However, Kimi K3 makes one thing abundantly clear: the US lead is continuing to shrink. Despite sanctions, Chinese AI labs are succeeding in training increasingly capable models. Whether reinforcement learning or distillation is the primary reason, I cannot say. But Kimi K3 demonstrates to the world that a turning point may be approaching. China is steadily moving toward technological parity. The statement by the founder of Zhipu AI should also attract attention: the company says it plans to release a GLM model in the Mythos class by the end of the year.

Kimi K3 is not Fable, and it is not Mythos. But it is already very close to the leading Western state-of-the-art models. That is highly significant.

[Kimi.ai @Kimi_Moonshot]

Introducing Kimi K3: Open Frontier Intelligence

🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal
🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts
🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost
🔹 Built for long-horizon agentic coding and self-evolving workflows

Kimi K3 is now live on on https://t.co/zrk6zZxZUo, Kimi Work, Kimi Code, and the Kimi API.
Open Weights by July 27, 2026.

🔗 API: https://t.co/XCrgjXAqMw
🔗 Tech blog: https://t.co/YTfiMSNM1f

[にこそく @nicosokufx]

中国メモリー大手ＣＸＭＴ、27日上海上場へ
今年アジア最大規模

昨日からのメモリー株の逆風材料の一つと言われている

[Pankaj Kumar @pankajkumar_dev]

Kimi K3 Launched

- Kimi K3 is now live the largest open-source MoE model yet with 2.8T parameters and a 1M context window.
- API pricing: $3/$15 per 1M tokens ($0.30 cache hits), around 5× more expensive than K2.7
- Benchmarks are impressive: #2 on AA-Briefcase, 91.2 on BrowseComp (SOTA), and behind only Claude Fable 5 Max & GPT-5.6 Sol Max on GDPval-AA v2.
- Frontend generation is exceptional. From my testing, it beats Opus 4.8 and is close to Fable 5 in UI taste.
- Built on a new KDA (Kimi Delta Attention) + AttnRes architecture with 896 experts (16 active/token) for highly efficient sparse routing.

[片山幹健（Tomotake Katayama) @tomotake94]

インドのAIコーディング企業のEmergentが、シリーズCで1.3億ドル（約208億円）を調達。
ポストマネー評価額は15億ドル（約2,400億円）に。

リードはPEのCreaegis。新規でMNI Ventures-Claypond、Sentinel Global、既存からKhosla Ventures、ソフトバンク・ビジョン・ファンド2、Lightspeed、Y Combinatorが参加。

2025年創業で設立約1年でのユニコーン到達。
ARRは1.2億ドル（約190億円）で直近4カ月で+70%、有料顧客は20万超。従業員は約200名。
ターゲットはエンジニアではなく、起業家とSMB（中小企業）で、トラック運送業の配送追跡、工場・建設業のERP、不動産管理の顧客管理など、メール・スプレッドシートで回してきた業務のアプリ内製化を担っています。
売上構成は北米約1/3、欧州約1/3、その他1/3で、母国インドは8〜9%。

AIコーディングの中でもポジショニングが明確で、直接の競合はReplitと明言しており、Claude Code・Codex・Cursorら開発者向けツールとは異なり、「非技術者にはデプロイ・ホスティング・テスト・デバッグまで面倒を見るプラットフォームが要る」というアプローチ。
また、コード生成ではなく業務システム内製の代行という価値提案で、ツール提供というよりIT受託の考え方に近い印象を受けました。

https://t.co/VNOylt1ZBH

[日本経済新聞 電子版（日経電子版） @nikkei]

メモリー市場「シリコンサイクル」波形から階段状に　需給逆転で新段階
https://t.co/L4DkgHnWNI

メモリー半導体市場は3〜5年で好不況を繰り返してきました。最近はAI需要でかつてない売り手市場に。メモリー市場は「スーパーサイクル（需要の急拡大期）に入っていく」のかーー。

[田中 洸輝 🏈 More than a VC @TANANY_VC]

創業13ヶ月でユニコーン、売上187億円
YC発のスタートアップ Emergentが、Series C 約200億円を調達、評価額は半年で約5倍の約2,350億円に

今もスプレッドシートとメール、FAX、電話で回っている中小企業（特に運送会社、工場、建設会社、不動産管理会社といった業界）に集中

コードを書く手助けをするのではなく、デプロイもホスティングもテストもデバッグも、全部引き受ける

有料顧客は20万を超え、190カ国で1,200万超のアプリが開発された

[BridgeMind @bridgemindai]

Kimi K3 just beat Fable 5 on the BridgeBench Horror House game test.

I did not expect this at all.

Kimi K3 is better than Fable 5 at game development and UI design.

The two things Fable 5 was supposed to own.

The 5x price increase suddenly makes sense.

[Angel 🌼 @Angaisb_]

My thoughts on Kimi K3 while we wait for the benchmarks

I think that it's mostly hype. In their blog they say it's better than Opus 4.8 but they don't show coding benchmarks, just those who are somewhat related to frontend, because that's what this model excels at

They also say it's cheaper than other models but we still don't know how efficient it is, sure, it's cheaper per M output tokens, but it's useless if it uses 5x as many tokens for the same task. So far from my testing efficiency isn't looking good

Is it a good model? Yes, all open-source models are welcome, but they're not useful for 99% of people, there has been too much hype

[Car Watch @car_watch]

日産、新型「エルグランド」は「e-4ORCE」の2グレード設定で価格は689万7000円から WLTC燃費は16.8km/Lを実現 https://t.co/koB87SlXsA #Nissan #ELGRAND

[たけぞう @noatake1127]

アップル、AI半導体企業の買収を模索

[荒井健一 AK @aarai666]

DeepSeek創業者の純資産が5.3兆円になった、創業わずか3年！

7月15日に更新されたブルームバーグ・ビリオネア指数の最新データによると、2023年7月に設立されたDeepSeekの創業者・梁文鋒（リャン・ウェンフォン）の純資産が360億ドル（約5.3兆円）になり、世界長者番付の63位にランクインした

生成AIのスタートアップの創業者としては、世界で最も高い資産を持つ人物となった

梁氏の現在の純資産は360億ドル（約5.3兆円）で、年初から199億ドル（約2.9兆円）増加。増加率は120％を超えた。

ちなみにDeepSeek は、2023年7月に設立された、わずか3年ほどの企業です。

皆さん、AIドリームの時代です