2026.04.27 AI News Briefing — Google unveils 8th-gen TPUs, Snap announces mass layoffs: AI infrastructure and job shocks at once

The AI industry moved on two fronts today. Google jumped fully into the infrastructure race by unveiling its 8th-generation TPUs aimed at the agentic AI era, while Snap announced large layoffs, saying 'AI writes 65% of our code'. In Korea, Samsung and SK hynix posted record results on AI chip demand.

---

1. Google unveils 8th-gen 'TPU 8t and 8i' — built for the agentic era

• Source: Google Cloud / Radical Data Science | 2026.04.22–24
• What happened: Google Cloud announced its 8th-generation TPUs at 'Google Cloud Next 2026'. Notably, they are split into two architectures: 'TPU 8t' for training and 'TPU 8i' for inference. The culmination of roughly a decade of development, they are designed to cover everything from next-generation model training to large-scale inference workloads. An 'agentic enterprise' vision was presented alongside an integrated AI stack.
• Why it matters: Big Tech's in-house chip race to reduce dependence on Nvidia GPUs is now in full swing. The cost structure of AI infrastructure itself is likely to be reshaped.

2. Snap: "AI writes 65% of our code" — announces about 1,000 layoffs

• Source: Crescendo AI News | 2026.04
• What happened: Snap CEO Evan Spiegel announced roughly 1,000 layoffs and the closure of about 300 open positions. The company explained that 'rapid advances in AI let us achieve the same results with fewer people', noting that more than 65% of new code is now AI-generated. The restructuring is expected to save more than $500 million a year by the second half of 2026. The stock rose 11% in after-hours trading right after the announcement.
• Why it matters: It is a case of a company publicly acknowledging that 'AI replaces people', moving past the stage of 'AI assists people'. It reads as the opening signal of a shift in the tech industry's employment structure.

3. Samsung Electronics and SK hynix post record Q1 results on the AI chip supercycle

• Source: Click Today / industry roundup | 2026.04.26
• What happened: Samsung Electronics and SK hynix recorded their best-ever results in Q1 2026. The main drivers were strong high-bandwidth memory (HBM) prices, rising demand for server DRAM, and growing demand for enterprise SSDs (eSSD). As AI spreads from training-centric to inference and agent workloads, demand for high-value memory is growing even larger. Samsung showed strength in portfolio diversification, SK hynix in its concentration on AI memory.
• Why it matters: The numbers confirm that the Korean economy is a direct beneficiary of the AI infrastructure build-out. Geopolitical risk in the Middle East remains a latent variable, however.

4. Meta unveils in-house model 'Muse Spark' — pivoting from open-source Llama to closed models

• Source: Crescendo AI News | 2026.04
• What happened: Meta unveiled 'Muse Spark', the first product of its newly formed Superintelligence Labs. It is the company's first closed model, led by Chief AI Officer Alexandr Wang, marking a turn away from the open-source Llama strategy. Meta says it delivers competitive performance in multimodal perception, reasoning, health, and agentic tasks while running at a fraction of the compute cost of the mid-sized Llama 4 model. Meta has set its 2026 AI capital expenditure at $115–135 billion.
• Why it matters: Meta has partly given up its identity as the 'open-source champion' and joined the closed-model race. It is an important inflection point that unsettles the center of gravity of the open-source camp.

5. Novo Nordisk × OpenAI integrate AI across all of drug development

• Source: Crescendo AI News | 2026.04
• What happened: Danish pharmaceutical company Novo Nordisk announced a strategic partnership with OpenAI. The core of the deal is integrating AI across the entire business, from drug discovery and clinical trials to manufacturing, supply chain, and commercial operations, with full deployment targeted by the end of 2026. By accelerating the discovery of obesity and diabetes drug candidates, it aims to regain the market leadership lost to Eli Lilly. The company stressed that 'AI is a tool that augments scientists rather than replacing them'.
• Why it matters: It is the first major case of global Big Pharma adopting AI as a 'corporate operating system' rather than an R&D aid. It could set a new benchmark for the pace of AI adoption in healthcare.

6. California accelerates AI legislation — key bills including SB 1000 up for review on April 27

• Source: Transparency Coalition | 2026.04.24
• What happened: AI-related bills are moving quickly through the California legislature. SB 1000, which covers AI disclosure and provenance data, faces an Appropriations Committee hearing on April 27, and many other bills are under review, including mandatory disclosure of AI use in advertising (SB 1050) and expanded penalties for AI digital replicas and deepfake extortion (SB 1015 and SB 1111). SB 928, which requires that instructors at California State University (CSU) be human, also passed committee.
• Why it matters: California legislation usually becomes the de facto standard across the US. Korean AI service providers are likely to face a longer pre-launch checklist.

---

Today's insight

• The in-house chip race for AI infrastructure begins in earnest: Google's 8th-generation TPU launch marks the start of a 'hardware sovereignty race' beyond the 'model race'. As the structure of sole dependence on Nvidia GPUs loosens, Korea's memory and foundry industries face new opportunities and risks at the same time.
• The AI job shock enters the 'acknowledged' stage: The Snap case is one of the first major instances of a company officially citing 'AI replacing jobs' as its reason. Redesigning roles and reskilling strategies will become a task for every company.
• The open-source vs. closed-model landscape is being redrawn: With even Meta pivoting to closed models, the center of the open-source camp is likely to shift toward DeepSeek, Mistral, and others. It is time to re-examine the changing map of model options when shaping a Korean LLM strategy.