Apple ships M6, WRTN becomes a unicorn, Gatik raises $200M, and more
Apples 2nm M6 chip, Intels 256-core Diamond Rapids, WRTN becoming Koreas first AI service unicorn, and Emerald AIs data center power optimization — nine stories from the last week of August.
AI Agents
Enterprise agent deployments surge into production
Per CrewAI’s 2026 survey, every surveyed enterprise plans to expand agentic AI use this year. Roughly 75 percent classify it as a core priority or strategic initiative. The median payback period on agent deployments is 5.1 months.
An Agentic AI Institute report finds 72 percent of enterprises have deployed to production while a 60 percent governance gap remains. Banking and insurance lead adoption at 47 percent. Healthcare (18 percent) and government (14 percent) are still early.
The gap between deployment speed and safety infrastructure is the central problem. Companies rushing without risk management face project cancellations. Clear ROI measurement and governance frameworks are the urgent needs.
Big Tech
Apple announces Mac Mini with the 2nm M6
Apple announced a new Mac Mini with M6 and M5 Pro chips. M6 is Apple’s first chip on a 2nm process, with a 12-core CPU and 12-core GPU.
According to Apple, AI performance is 4x and graphics 2x that of M4. CPU performance is up 40 percent and storage is twice as fast. The M6 model starts at $899, the M5 Pro at $1,699.
Wi-Fi 7, Bluetooth 6, and 2.5Gb Ethernet are standard. It ships 22 September. Commercializing a 2nm process is a milestone in the on-device AI performance race.
Alibaba Cloud opens its first South American data centers in Brazil
Alibaba Cloud opened two data centers in São Paulo, Brazil — its first cloud region on the South American continent, following the Mexico region in February 2025.
It plans to launch its AI agent service suite in Brazil. The company now operates 31 regions and 106 availability zones worldwide, targeting local enterprises, startups, and public institutions.
There are also concerns about a collision between China’s intelligence law and Brazil’s LGPD. Data sovereignty is a recurring obstacle in Chinese cloud providers’ global expansion.
Intel Xeon 7 Diamond Rapids, 256 cores, arriving 2027
Intel revealed its next-generation server CPU, Xeon 7 Diamond Rapids, with up to 256 P-cores in a single socket, PCIe 6.0, and 16-channel memory.
Built on Intel 18A-P, it delivers 9 percent more performance at the same power. A 512-core model follows within months of launch. Memory bandwidth doubles over the previous generation.
Originally planned for 2026, it slipped to 2027. AMD’s EPYC Venice is expected to ship first. Diamond Rapids is the last Xeon generation without hyperthreading.
Startup Funding
Gatik raises a $200M Series D for driverless freight
Autonomous middle-mile freight startup Gatik raised a $200 million Series D led by the Qatar Investment Authority and Koch Disruptive Technologies, bringing total funding to roughly $500 million.
It has completed 85,000 driverless deliveries on Walmart, Kroger, and PepsiCo routes. Contracted revenue exceeds $600 million, with a 99 percent on-time delivery rate.
Its strategy of focusing on the middle mile between distribution centers and stores is paying off. The target is operating more than 100 driverless trucks by the end of 2026 — a company with a proven record in autonomous freight.
WRTN raises ₩100B Series C, becoming a Korean AI unicorn
Wrtn Technologies raised approximately ₩100 billion in its Series C, passing a ₩1 trillion valuation — the first unicorn among Korean AI service startups.
Its AI entertainment platform OOC reached ₩10 billion in monthly revenue three months after launching in North America in May. The company is targeting ₩200 billion in revenue this year. Cumulative funding is ₩230 billion.
The notable detail is overseas revenue overtaking domestic. It demonstrates global growth in a new category — AI entertainment — and stands as a success case for Korean AI startups expanding abroad.
Emerald AI hits unicorn status on a Series A for data center power
Emerald AI raised $150 million in its Series A, reaching a $1.05 billion valuation. NVIDIA, Samsung Ventures, and Siemens participated.
Its Conductor platform coordinates AI workloads with energy resources, dynamically controlling data center power consumption during grid stress. It targets more than 100GW of underutilized grid capacity in the US.
It is a critical technology as AI data center power demand surges. Power shortage is the biggest bottleneck in AI infrastructure expansion, and approaching it in software is what sets the company apart.
Open Source
OpenViking: a context database for AI agents
OpenViking, released by ByteDance’s cloud infrastructure
arm Volcengine, has passed 30,000 GitHub stars. It manages an agent’s memory, knowledge, and skills
under a filesystem paradigm, with agents navigating context through a viking:// protocol.
Content is processed at three levels — L0 (abstract), L1 (outline), L2 (detail) — loading only the level needed, on demand. Unlike conventional vector stores, it uses a filesystem metaphor.
The underlying paper, VikingMem, was accepted at VLDB 2026. It is an academically validated approach to agent memory management, with over 2,300 forks in active use.
MoneyPrinterTurbo: automated AI short-form video generation
MoneyPrinterTurbo has passed 118,000 GitHub stars. It generates HD short-form video from a single keyword, automating script, footage, subtitles, voice, and background music.
It supports direct upload to TikTok, Instagram Reels, and YouTube Shorts, and offers a WebUI, REST API, CLI, and an AI agent interface. Docker deployment keeps installation simple.
The MIT license permits commercial use freely, and it supports a range of LLM and TTS providers. It is the most actively developed open-source project among content creation automation tools.