AI in Asia
Sunday, 19 April 2026

3 Before 9

3 must-know AI stories before your 9am coffee

Who should pay attention

Enterprise AI buyers | Startup founders | Policy makers | Cloud providers | Compliance teams

What changes next

Expect AI governance to tighten across ASEAN as Vietnam's binding AI law sets the precedent for the region.

1

PwC Study Finds 74% of AI Value Captured by Just 20% of Companies

PwC released its 2026 AI Performance Study this week, surveying 1,217 senior executives across 25 sectors worldwide and finding that nearly three-quarters of AI's economic gains are flowing to roughly one-fifth of organisations. The leading companies share a common trait - they are using AI primarily to pursue growth rather than just cut costs. Growth-focused firms are two to three times as likely as their peers to deploy AI for identifying opportunities across industry boundaries, such as partnering with companies outside their core sector. The study found that efficiency-only strategies deliver diminishing returns, while companies combining AI with cross-sector expansion are pulling sharply ahead.

Why it matters for Asia

Singapore stands out in the data. Sixty-seven per cent of Singapore businesses say they are willing to take risks when investing in AI, compared with 41 per cent globally, and 43 per cent of Singapore respondents report using AI to compete with companies outside their own industry versus just 20 per cent worldwide. For enterprise buyers across Asia-Pacific, the message is clear - the AI advantage now belongs to companies chasing new revenue, not those trimming headcount, and Singapore-based firms appear to be betting accordingly.

Read more
2

Google Releases Gemma 4 Under Apache 2.0 With Frontier-Class Performance

Google DeepMind released Gemma 4 in early April, a family of four open-source models ranging from a compact 2B-parameter edge variant to a 31B dense model that currently ranks third on the Arena AI text leaderboard. All four models can understand images and video as input, with the smaller variants also accepting audio for speech recognition - though it is important to note that Gemma 4 only generates text output, not images, audio, or video. The 31B model runs entirely on a single 80GB Nvidia H100 GPU while matching or exceeding benchmarks set by proprietary models many times its size. The entire family ships under a fully permissive Apache 2.0 licence, meaning unrestricted commercial use, modification, and redistribution with no royalty obligations.

Why it matters for Asia

For Asian startups and enterprise teams that cannot afford frontier API pricing or prefer to keep data on-premises, Gemma 4 changes the calculus. A model that fits on one GPU, speaks multiple languages natively, and carries no licensing restrictions removes a major barrier for developers in markets like India, Indonesia, and Vietnam where cost sensitivity and data sovereignty concerns run high. Expect regional cloud providers across Southeast Asia and Japan to move quickly to offer hosted Gemma 4 instances as a competitive wedge against OpenAI and Anthropic.

Read more
3

Southeast Asia Leads World in AI Optimism While Governance Frameworks Lag Behind

Stanford University's 2026 AI Index Report and a new survey highlighted by TechWire Asia show that Southeast Asian populations are the most optimistic about AI anywhere in the world, but the region's governance infrastructure has not kept pace. In Malaysia, Thailand, Indonesia, and Singapore, more than 80 per cent of respondents say AI will profoundly change their lives within three to five years. Singapore leads all 30 surveyed countries at 81 per cent public trust in government AI regulation, followed by Indonesia at 76 per cent, Malaysia at 73 per cent, and Thailand at 70 per cent. Yet most ASEAN nations still rely on voluntary frameworks and soft-law guidance rather than binding AI-specific legislation, with Vietnam the sole exception after enacting an AI law in December 2025 that takes effect in phases from March 2026.

Why it matters for Asia

The gap between public enthusiasm and regulatory readiness creates both opportunity and risk for companies deploying AI across the region. Businesses moving fast in markets with minimal binding rules enjoy short-term flexibility, but face the prospect of retroactive compliance costs when hard regulation inevitably arrives. Vietnam's early move signals the direction of travel, and enterprise buyers building AI products for Southeast Asian markets should be stress-testing their governance playbooks now rather than waiting for each country to follow suit.

Read more

That's today's 3 Before 9.

Explore more at AIinASIA.com or share signals with us.

Recent Editions

View all

Friday

24 April 2026

  • 1. DeepSeek launches V4-Pro and V4-Flash with a 1-million-token context window, running on Huawei's new Ascend 950 supernode clusters and priced at roughly a tenth of OpenAI's output-token rate.
  • 2. SoftBank is seeking a $10 billion two-year margin loan collateralised by its OpenAI stake, pushing its total commitment to the ChatGPT maker to about $64.6 billion.
  • 3. TSMC unveiled A13, a 1.3nm-class shrink of A14 aimed at AI accelerators, with production slated for 2029 alongside a 1.2nm A12 variant, while A16 slips to 2027.
Read edition

Thursday

23 April 2026

  • 1. Microsoft pledges A$25 billion (US$17.9 billion) to make Australia its largest AI infrastructure hub, including skills training for three million workers.
  • 2. SK Hynix posts a record 72 per cent operating margin and US$27 billion quarterly profit as HBM demand keeps Asia's memory duopoly in the driver's seat.
  • 3. Japan's Nikkei 225 breaches 60,000 for the first time, but only 17 per cent of Tokyo stocks rose on the day, exposing a dangerously narrow AI-driven rally.
Read edition

Wednesday

22 April 2026

  • 1. LINE Yahoo launched Agent i on 20 April, a single AI agent now embedded in both the LINE app's 100m-plus user base and Yahoo! JAPAN, with enterprise and no-code versions due over the summer.
  • 2. SoftBank named Arm CEO Rene Haas to also run SoftBank Group International from 21 April, tightening coordination across the group's overseas chip and AI bets including Arm, OpenAI, Stargate and Project Crystal Land.
  • 3. Meta's $2 billion acquisition of Singapore-based Chinese AI start-up Manus is still under Beijing's national-security review, with co-founders barred from leaving China and the 'Singapore washing' route for China AI firms now under serious pressure.
Read edition

Tuesday

21 April 2026

  • 1. Qualcomm CEO Cristiano Amon met Samsung and SK Hynix executives in Seoul to pitch 2nm wafer orders and lock in memory supply for Snapdragon and AI platforms.
  • 2. Chinese humanoid robot makers shipped roughly 90 per cent of global units in 2025 but still trade at a valuation discount to pre-production US rivals.
  • 3. Volkswagen confirmed its China cars will ship with an on-device AI voice agent from H2 2026, built on tech from Tencent, Alibaba and Baidu.
Read edition

Monday

20 April 2026

  • 1. SK Hynix has begun mass production of its 192GB SOCAMM2 memory module, a low-power AI server part for Nvidia's Vera Rubin platform.
  • 2. Alibaba launched Qwen3.6-Max-Preview, its strongest Qwen model to date, demonstrating top performance in several coding benchmarks.
  • 3. TSMC raised its 2026 outlook after Q1 profit jumped 58 per cent, citing "extremely robust" AI chip demand that still outpaces supply.
Read edition

Saturday

18 April 2026

  • 1. TSMC reported a 58% increase in Q1 net profit, driven by record demand for AI chips and advanced node sales, leading to higher full-year revenue guidance and global expansion.
  • 2. Alibaba introduced "Happy Oyster", an AI world model that generates consistent, interactive 3D environments in real time, aiming to challenge Tencent in film, gaming, and VR applications.
  • 3. South Korean president Lee Jae Myung arrives in New Delhi for a state visit, with a Monday summit set to cover AI, shipbuilding, defence and small modular reactors alongside a $50 billion bilateral trade target.
Read edition