{"slug": "neuralwatt-doubling-the-price", "title": "Neuralwatt Doubling the Price", "summary": "Neuralwatt Cloud will double its standard energy rate from $5/kWh to $10/kWh on July 16, 2026, citing the need to support growing scale and invest in reliability and capacity. The company claims energy-based pricing remains cheaper than token-based pricing for most workloads and is introducing tiered pricing for latency-tolerant inference. Existing credits and subscriptions are unaffected, but $1,000 PAYG packs and annual subscriptions are temporarily unavailable.", "body_md": "# Pay for Compute, Not Token Markups\n\nEnergy-based pricing aligns your costs with actual GPU resources consumed.\n\n## New pricing rolling out July 16\n\nWe've been thrilled by the response to Neuralwatt Cloud and the launch of flex inference.\nTo support our growing scale and invest in reliability, capacity, and new pricing tiers,\nwe're updating our standard energy rate from **$5/kWh to $10/kWh** on\n**July 16, 2026**. Energy pricing will still be substantially cheaper than\ntoken-based pricing for most workloads, and the efficiency compounds with flex.\n\n### What's changing July 16\n\n-\nBase energy rate:\n**$5 → $10/kWh** - PAYG packs become flexible: buy any amount from $5–$1000 at the flat rate\n- Subscription tiers adjusted for the new rate\n\n### Temporary (until July 16)\n\n-\n**$1,000 PAYG packs** and**annual subscriptions** are temporarily unavailable while we prepare the new pricing - Monthly subscriptions, $10–$50 packs, and all existing credits remain fully available\n- Your current credits and subscriptions are unaffected\n\n### What's coming: pricing tiers by flexibility\n\nWe're building tiered pricing that lets you (and your agents) optimize price vs. performance depending on how much latency you can tolerate. The following are early examples. Exact discount levels and availability are still being finalized.\n\nLatency-tolerant inference at a discount. Greater hold tolerance = deeper savings.\n\nSynchronous inference with improved SLA and throughput guarantees.\n\nPriority lanes with tighter TTFT and throughput SLAs, billed at a premium.\n\nDedicated capacity, sovereign data residency, higher concurrency, tighter SLAs.\n\nWe're also working on API access to real-time energy/request data, flex hold times per discount level, and throughput metrics, so your agents can self-optimize price/performance. More detail to come as these ship.\n\nWe're grateful for your engagement and support as we grow. If you have questions, reach out in our\n[Discord](https://discord.gg/ZJEfU2BZw2)\nor email\n[[email protected]](/cdn-cgi/l/email-protection#640d0a020b240a0111160508130510104a070b09).\n\n## Monthly Subscription Plans\n\nPredictable monthly pricing with included energy allocation. Know exactly what you'll pay each month.\n\n### Basic\n\nPerfect for personal projects and light AI workloads.\n\n- Access to all models\n- Real-time energy tracking\n- Energy dashboard\n\n### Standard\n\nGreat value for everyday AI productivity with priority access.\n\n- Everything in Basic\n- 2.7x more energy than Basic\n- Priority access during high demand\n- Usage analytics\n\n### Pro\n\nFor power users with the most energy and highest priority.\n\n- Everything in Standard\n- 5.5x more energy than Basic\n- Highest priority access\n- Overage billed at your plan's discounted rate (once your plan renews after July 16)\n\n* As part of the [July 16 pricing update](#pricing-update),\nnew allowances reflect the $10/kWh rate with your tier's discount (2% Basic, 3% Standard, 5% Pro) and apply to\nsubscriptions that **start or renew on or after July 16**. A subscription bought today keeps its\ncurrent allowance for its full term; nothing changes mid-cycle.\n\n#### Why Subscribe?\n\nSubscriptions are billed in kWh (energy), not tokens. This means the same rate applies to all models: use a small, fast model or a large reasoning model with zero markup. You always know exactly what you'll pay.\n\n## API Access & Pay As You Go\n\nNo commitment required. Start with $1.00 free credits, then add more when you need them.\n\n### Energy-Based Pricing\n\nPay for actual compute consumed\n\nPay for actual GPU compute consumed, not arbitrary per-token markups. One rate for all models: efficient MoE models cost dramatically less than traditional token pricing.\n\n- One flat rate, no per-model markups\n- MoE models (Qwen3.5, Kimi K2.5) up to 95% cheaper\n- Efficient prompts cost less, so optimization pays off\n- Full cost transparency\n- Prefix caching: cached tokens skip recomputation, reducing energy vs full prefill\n\n### Token-Based Pricing\n\nAlternative for familiar billing\n\nStandard per-token billing if you prefer familiar pricing. Rates vary by model; consider energy pricing for better value on most models.\n\n- Familiar per-token model\n- Easy migration from other providers\n- Energy reporting included free\n\n#### Free Trial\n\n$1.00 credit included, 5 req/min, 50 req/day\n\n#### Pay As You Go\n\n$10 minimum, 500 req/min, unlimited daily\n\n### Model Pricing\n\nToken prices vary by model. Energy pricing uses a flat **$5.00/kWh** for all models, dramatically cheaper for efficient models.\nThe energy figure below is the **average measured energy at a typical request size** from real usage over the trailing 7 days, not a fixed rate; energy scales with request size.\n[Understand energy pricing & compare costs →](/energy-pricing)\n\n| Model | Input |\nCached Input†\n|\nOutput | Avg Energy/Req | |\n|---|---|---|---|---|---|\n|\nGLM-5.2\nZhipuAI\n|\n$1.45/M tokens | $0.36/M tokens | $4.50/M tokens | ~1.97 Wh |\n|\n\n[Try](/playground/glm-5.2-fast)[Try](/playground/glm-5.2-short)[Try](/playground/glm-5.2-short-fast)[Try](/playground/kimi-k2.6)[Try](/playground/kimi-k2.6-fast)## Team & Enterprise\n\nDedicated infrastructure, custom SLAs, and volume discounts for organizations.\n\n### Enterprise\n\nFor organizations that need dedicated GPU infrastructure, guaranteed capacity, and enterprise-grade support.\n\n[Contact Sales](/cdn-cgi/l/email-protection#771e1911183719120205161b001603035914181a)\n\n### Neuralwatt Deploy\n\nBring Neuralwatt's energy optimization directly into your data center. The same Neuralwatt Optimize engine powering Neuralwatt Cloud, deployed on your infrastructure.\n\n**Full control over your hardware, security, and power consumption.** We'll help you\ndeploy the same optimizations powering this service.\n\n## All Models\n\nToken-based pricing varies by model. Energy-based pricing uses a flat **$5.00/kWh** rate for all models, up to 95% cheaper on efficient architectures.\n\n[How we measure energy](/docs/energy-methodology)\n\n†: $0.36/M tokens\n\n[Try Now](/playground/glm-5.2)\n\n†: $0.36/M tokens\n\n[Try Now](/playground/glm-5.2-fast)\n\n†: $0.36/M tokens\n\n[Try Now](/playground/glm-5.2-short)\n\n†: $0.36/M tokens\n\n[Try Now](/playground/glm-5.2-short-fast)\n\n†: $0.17/M tokens\n\n[Try Now](/playground/kimi-k2.6)\n\n†: $0.17/M tokens\n\n[Try Now](/playground/kimi-k2.6-fast)\n\n†: $0.24/M tokens\n\n[Try Now](/playground/kimi-k2.7-code)\n\n†: $0.17/M tokens\n\n[Try Now](/playground/qwen3.5-397b)\n\n†: $0.17/M tokens\n\n[Try Now](/playground/qwen3.5-397b-fast)\n\n†: $0.07/M tokens\n\n[Try Now](/playground/qwen3.6-35b)\n\n†: $0.07/M tokens\n\n[Try Now](/playground/qwen3.6-35b-fast)\n\n| Model | Context | Token Input |\nCached Input†\n|\nToken Output | Avg Energy/Req* | Features | |\n|---|---|---|---|---|---|---|---|\n|\nGLM-5.2\nZhipuAI\n|\n1048K | $1.45 /M tokens | $0.36 /M tokens | $4.50 /M tokens | ~1.97 Wh |\nR\nT\n|\n|\n\n[Try](/playground/glm-5.2-fast)[Try](/playground/glm-5.2-short)[Try](/playground/glm-5.2-short-fast)[Try](/playground/kimi-k2.6)[Try](/playground/kimi-k2.6-fast)[Try](/playground/kimi-k2.7-code)[Try](/playground/qwen3.5-397b)[Try](/playground/qwen3.5-397b-fast)[Try](/playground/qwen3.6-35b)[Try](/playground/qwen3.6-35b-fast)\n*Average measured energy at a typical request size from real usage over the trailing 7 days, pooled across all servers, not a fixed rate. Energy scales with request size; see the full grid.\n[Learn more](/energy-pricing)\n\n†Cached input tokens are billed at **25% of the input rate**. Prefix caching makes long-context requests substantially cheaper when the same prefix is reused. The 25% is the default rate; per-model rates may be set in future, in which case the per-model rate applies. This same effective rate is advertised as `cached_input_per_million`\n\non [/v1/models](/docs/api/models).\n\n## Frequently Asked Questions\n\n**Energy pricing** is recommended for most users; it's dramatically cheaper for efficient MoE models (up to 95% savings vs token pricing) and gives you full cost transparency.\n\n**Subscriptions** combine energy pricing with a monthly discount (~35% off PAYG rates).\n\n**Token pricing** is available as a familiar alternative if you prefer per-token billing. All options include free energy reporting.", "url": "https://wpnews.pro/news/neuralwatt-doubling-the-price", "canonical_source": "https://portal.neuralwatt.com/pricing", "published_at": "2026-07-10 05:34:29+00:00", "updated_at": "2026-07-10 06:06:28.459902+00:00", "lang": "en", "topics": ["ai-infrastructure", "ai-products", "ai-startups"], "entities": ["Neuralwatt Cloud"], "alternates": {"html": "https://wpnews.pro/news/neuralwatt-doubling-the-price", "markdown": "https://wpnews.pro/news/neuralwatt-doubling-the-price.md", "text": "https://wpnews.pro/news/neuralwatt-doubling-the-price.txt", "jsonld": "https://wpnews.pro/news/neuralwatt-doubling-the-price.jsonld"}}