Data Hub ▍Exclusive data & original research

Chunghwa Telecom, Taiwan Mobile, FarEasTone — H1 2026 revenue, profit, EPS and 5G compared. Notably Taiwan Mobile leads on EPS (NT$2.86), but partly via its momo e-commerce and non-operating items; FarEasTone has the industry's highest 5G penetration.

The world's largest sovereign fund, Norway's NBIM, and its AI-core holdings in NVIDIA, TSMC, Apple, Microsoft and more — plus the verifiable US positions of Temasek, PIF and others. NVIDIA is now NBIM's single largest holding.

US CHIPS, EU Chips Act, Japan/Rapidus, Korea, China Big Fund, Taiwan, India ISM — seven major chip-subsidy programs compared. Key: the headline numbers use wildly different bases (authorized ≠ disbursed ≠ registered capital) — read the basis before comparing.

Groq, Cerebras, SambaNova, AWS Trainium/Inferentia, Google TPU, Intel Gaudi — the AI-accelerator battlefield outside NVIDIA. Compared by architecture, memory type (large SRAM vs HBM) and deployment, highlighting differentiated paths like deterministic dataflow and wafer-scale single-chip.

Ollama, LM Studio, llama.cpp, vLLM, MLX, SGLang — 8 mainstream local/on-device LLM inference tools compared: positioning, platforms, model formats, license and use case. Four classes: desktop app / inference engine / Apple-native / production serving.

Phi-4, Gemma 3, Qwen3, Llama 4, DeepSeek V4, Kimi K2 — 8 mainstream open models × Apple Silicon, computing the minimum Mac + memory + estimated speed for inference. Key: MoE models make Macs punch above their weight, but 1T-class flagships exceed Mac memory limits even at Q4 — they simply won't run locally.

OpenAI, Anthropic, xAI, Perplexity, Mistral, Cohere — revenue, valuation and IPO status for the major AI companies. Key caveat: most headline "revenue" figures are annualized run rates, not realized annual revenue.

Tesla, BYD, Volkswagen, Kia — Q2 2026 pure-electric (BEV) deliveries. On a like-for-like "pure BEV, quarterly" basis, BYD's ~557k BEVs still lead the world, with Tesla second at 480k.

AWS, Azure and Google Cloud — Q2 2026 market share in global cloud infrastructure services (IaaS/PaaS). The big three hold ~67%; quarterly spend passed $143.4bn, up 43% (an eight-year high) driven by AI.

Apple, Microsoft, Alphabet, Amazon, NVIDIA, Meta and Tesla — latest market cap, quarterly revenue and profit for the "Magnificent 7", with NVIDIA on top at ~$5.2 trillion.

Fed, ECB, BOJ, BOE, Taiwan's CBC, China's PBOC and BOK — current policy rates and latest decisions across seven major central banks. In 2026 several turned hawkish or hiked as Middle East energy shocks pushed up inflation.

Fubon, Cathay, Nan Shan, Shin Kong, Taiwan Life, KGI and Mercuries — H1 2026 self-reported net income for Taiwan's seven big life insurers, together earning ~NT$180bn (up over 3x YoY) at record levels.

Fubon, Cathay, CTBC, Yuanta, KGI, E.SUN, Mega, First, Hua Nan and Taiwan Cooperative — H1 2026 self-reported net income and EPS for Taiwan's ten largest financial holding companies, with each group's main profit engine (life insurance / banking / securities).

MediaTek, Novatek, Realtek, Phison, GUC and Alchip — Q2 2026 consolidated revenue and profitability for Taiwan's six major fabless IC designers. Sharp divergence: Phison (+279%) and GUC (+128%) driven by AI storage and AI ASIC; MediaTek, Novatek and Realtek grew modestly.

H100, H200, H20, Blackwell, AMD MI series — a chip-by-chip map of US export controls on AI chips to China, disentangling three commonly-confused mechanisms: rules in force, the uncodified 15% revenue-share, and the rescinded AI Diffusion Rule.

NVIDIA's $500B financing platform, the OpenAI Ohio data-center backstop, Broadcom's AI-chip debt — a tracker of 2026's big AI-infrastructure financing and investment deals, with the widely-misreported figures calibrated (target vs in-talks vs signed).

LINE Bank, NEXT Bank and Rakuten — an H1 2026 operating scorecard for Taiwan's three pure-play internet banks. Combined accumulated losses narrowed to NT$6.94bn, but the three diverge sharply: LINE Bank turned operationally profitable, NEXT's losses widened, and Rakuten's drop is an accounting write-off.

Hon Hai, Quanta, Wistron, Wiwynn and Inventec — Q2 2026 consolidated revenue and AI-server momentum for Taiwan's five big AI-server ODMs, the supply-chain core of the global AI compute buildout.

Latest quarterly revenue and YoY growth for the three big memory makers — SK Hynix, Samsung, Micron — all at records amid the AI-driven memory super-cycle.

Latest quarterly revenue and YoY growth for the five big semiconductor equipment makers — ASML, Applied Materials, Lam Research, Tokyo Electron, KLA — the "picks and shovels" of the chip industry.

Latest valuations, funding rounds and lead investors for the major AI startups — OpenAI, Anthropic, xAI, Mistral, Perplexity and more — with a confidence tier per figure.

July 2026 consolidated revenue and YoY growth for Taiwan-listed AI hardware supply-chain companies — server ODMs, thermal, power, PCB/substrate, IC design, connectors — ranked by revenue.

Q1 2026 market share of the three HBM makers — SK Hynix, Samsung, Micron — plus HBM4 qualification status on Nvidia Vera Rubin, the key memory bottleneck for AI accelerators.

Q1 2026 revenue market-share ranking of the top-10 semiconductor foundries — TSMC, Samsung, SMIC, UMC and more — as AI demand pushes total top-10 revenue to a record.

Latest-quarter data-center / AI revenue, year-over-year growth and share of total revenue for NVIDIA, AMD, Intel and Broadcom — first-party earnings data connecting AI-accelerator specs to the money.

Current flagship models from OpenAI, Anthropic, Google, Meta, xAI, DeepSeek and Mistral compared on release date, context window, open weights, modalities, knowledge cutoff and license — every figure sourced to the provider's own official pages.

Latest-quarter capital expenditure, year-over-year growth and full-year guidance for Microsoft, Alphabet, Meta, Amazon and Oracle — first-party earnings data sizing the AI-infrastructure buildout.

Ten major AI/LLM capability benchmarks (MMLU, GPQA, SWE-bench, HumanEval, AIME, etc.) compared on what they measure, format/size, scoring and whether they are saturated — to make sense of the scores models claim.

Advanced process nodes from the big three foundries (TSMC, Samsung, Intel): mass-production timing, transistor architecture and representative use, ordered by year. Q4 2025 is where all three converge on the 2nm generation.

Major AI video generators (Runway / Luma / Pika / Google Veo) compared on free plan, entry paid tier, capability and commercial-use rights. Entry subscriptions run $8–15/month, while free plans generally forbid commercial use.

What it costs to run one NVIDIA H100 / H200 GPU per hour — the big three clouds (AWS / Azure / GCP) vs specialist GPU clouds (RunPod / Together / Lambda / CoreWeave), on-demand. The same GPU can cost 3–4x more depending on provider.

Major AI coding tools (GitHub Copilot, Cursor, Claude Code, Amazon Q, Replit, JetBrains, Tabnine, Windsurf) compared on free plan, personal paid price, underlying models, form factor and standout feature.

Specs across HBM generations (HBM2 / HBM2E / HBM3 / HBM3E / HBM4): per-pin data rate, per-stack bandwidth, max stack height and capacity. AI accelerator memory bottlenecks tie directly to HBM generation.

Nine popular Taiwan ETFs (market-cap / high-dividend / tech) compared on tracked index, management fee, size, distribution frequency, trailing-12-month cash yield and 1-year total return. Yield and total return side by side — a high yield does not mean higher returns.

Comparison of major data-center AI accelerators (NVIDIA / AMD / Google / Intel / AWS): memory, bandwidth, compute, power and cloud rental price. Specs from official documents, retrieved 22 July 2026.

Cross-vendor comparison of standard API pricing (input/output per 1M tokens) and Artificial Analysis capability scores for major LLMs. All figures sourced, retrieved 22 July 2026.