Taxation

1. Context and Rationale

The Australian Renaissance Party (ARP) recognises that compute is the primary driver of modern productivity and the direct competitor to human labour. As the "Labor Share of Capital" shifts toward silicon, the traditional tax base (payroll and income) is inherently threatened. To preserve social cohesion and ensure the continued relevance of human employment, the tax base must shift to include the "synthetic labour" that replaces it.

2. Approaches to Compute Taxation

Several approaches to taxing compute have been proposed or implemented, generally targeting one of three areas: the hardware (GPUs), the activity (energy/hashrate), or the output (tokens/inference). We have identified three primary candidates for compute taxation. Our position favours a hybrid model that prioritises administrative simplicity and national sovereignty, while recognising the need to explore taxation models that capture the value created by compute-intensive activities while avoiding stifling innovation.

• "Token Tax" & Inbound Intangibles: Taxing "token generation" or "inference services." Economists and AI safety researchers advocate for taxing "token generation" or "robot services" to fund a "Social Transition Fund" or Universal Basic Income (UBI) as automation scales. This addresses offshore providers (e.g., OpenAI, Anthropic, et al) who extract value from the Australian economy without maintaining local payrolls. This would be legislatively collected in the same way GST is collected now. Examples include the "AI Horizon Fund" proposal in the U.S. and Senator Sanders' advocacy for excise taxes on automating technology.
• Compute-Capable Hardware Tax (Input Excise): A levy on the "potential labour displacement" of hardware. This targets the physical entry of GPUs, NPUs, and ASICs at the border.
• The "Technical" and "Regulatory" Floor: This model involves setting a "floor" of public-interest obligations (e.g., safety audits, contributing compute to public research). Companies exceeding their "compute quota" or failing to meet their social "floor" would pay a levy or trade credits, utilising a Floor and Trade mechanism where companies meet a "Human Employment Ratio" or pay into a Social Integration Fund. This mirrors carbon markets or "charity-care" floors in healthcare. In engineering, a "Technical Compute Tax" may refer to the irreducible overhead of certain architectures, such as the quadratic compute tax in standard Transformer architectures.

3. Challenges and Considerations

The primary barrier to a formal compute tax is the "Innovation Paradox." Taxing compute could be seen as a "tax on the future," potentially driving domestic AI labs to migrate to jurisdictions with cheaper electricity and lower taxes, undermining national AI sovereignty.

Given these considerations, the ARP recognises the need to explore taxation models that capture the value created by compute-intensive activities while avoiding stifling innovation. The following principles should guide the development of any compute-related tax policy:

• Do No Harm to Innovation: Any taxation regime must not create incentives for AI labs and compute-intensive industries to relocate outside of Australia. By exempting legacy compute and providing export rebates, we ensure AI labs remain in Australia to serve the global market.
• Capture Externalities: Where compute-intensive activities generate negative externalities (e.g., excessive energy consumption), a "polluter pays" model may be appropriate.
• Share the Automation Dividend: As automation displaces human labour, mechanisms to capture a portion of the resulting economic gains for the benefit of all Australians should be explored. This could take the form of a levy on specific outputs or activities, carefully designed to minimise distortion.
• Pragmatic Regulation: We favour Input Taxes over Usage Taxes for local hardware to avoid the "Panopticon" of state-monitored processing.
• Human Centricity: We treat human employment as a necessary form of social integration. The tax system must be used to ensure that a booming AI economy does not result in a hollowed-out society.
• Sovereign Stockpiling: The "announcement effect" of this tax is a feature, not a bug. It will trigger a massive influx of compute capacity into Australia before the policy commences, providing the productivity backbone for our future GDP.

Further research and consultation are required to determine the most effective and equitable approach to taxing compute in Australia. This should include consideration of international best practices, potential economic impacts, and the need to maintain Australia's competitiveness in the global AI landscape. Note: Further research is required to define the "Industrial-Grade" threshold for hardware (e.g., TFLOPS/Watt) to ensure consumer devices like gaming consoles and mobile phones remain unaffected. This policy specifically addresses elements of our taxation model that related to compute-intensive activities, rather than general taxation principles. We are advocating a compute tax as it is compute that is directly competitive to human labour. While no nation has implemented a standalone, broad-based "Compute Tax Act" akin to a Goods and Services Tax (GST) or Income Tax, the concept is rapidly evolving to serious policy consideration in other countries. Attempts to tax compute generally target one of three areas: the hardware (GPUs), the activity (energy/hashrate), or the output (tokens/inference).

4. The ARP Integrated Model

The ARP advocates for a Circular Compute Tax model designed to fund human utility while making Australia a global "Compute Refinery."

A. The Hardware Input Tax (The "Anchor")

To solve the Edge Measurement Problem—the technical difficulty of metering compute in decentralised robotics or offline services, taxation shall be applied at the point of import or commissioning.

• Mechanism: An excise duty on high-density compute hardware (e.g., NVIDIA H-class/B-class or equivalent).
• Edge Advantage: Once the input tax is paid, the marginal cost of innovation is zero. This respects corporate privacy and eliminates the need for invasive "operational scales" or government metering of private code.
• Legacy Exemption: All hardware pre-purchased or commissioned before the policy enactment date is fully exempt. This avoids sovereign risk and encourages an immediate "Pre-Policy Rush" to stockpile infrastructure.
• Safety Exemption: All hardware deployed for safety related functions is fully exempt where it is deployed in safety subsystems where its sole responsibility is safety.

B. Inbound Token-GST

Following the existing GST framework for digital services, offshore compute providers must collect a levy on all tokens or inference cycles sold to Australian residents.

• B2B Reverse Charge: Australian businesses using offshore APIs will self-assess this levy, ensuring a level playing field for domestic compute providers.

C. The Export Rebate (The "Refinery" Incentive)

To encourage Australia to become a global hub for AI training and inference:

• Any compute tax paid on hardware or tokens is refundable if the resulting output (tokens, trained models, or data services) is exported to international clients.

5. The Human Utility Mechanism

Revenue from compute-intensive taxation is not directed to general revenue, but is strictly earmarked for the Social Integration Fund.

The Negative Payroll Tax

The primary destination for compute tax revenue is a Negative Payroll Tax. This acts as a direct subsidy for human employment this coupled with a Floor & Trade mechanism uses capitalism to find the least inefficient employment of labour by distributive search as per Hayek.

• Goal: To lower the "cost-to-employer" of a human worker without reducing the worker's take-home pay.
• Formula: As compute density increases, the negative payroll tax scales to make human labour mathematically competitive with automated alternatives in high-social-value sectors (e.g., Agriculture, Specialised Mining, Healthcare).

The GDP Inverse Relationship

The ARP recognises that "Tax buys civilisation." However, the burden of this tax is inversely proportional to national wealth:

$$T_c \propto \frac{\text{Social Integration Cost}}{\text{National GDP}}$$

As the ARP’s expansion of the Mining and Agricultural sectors drives up total GDP, the required levy on compute decreases. A high-GDP Australia can afford a "light-touch" compute tax while still fully funding the human employment floor.

6. ARP Position (Provisional)

• Do No Harm to Innovation: By exempting legacy compute and providing export rebates, we ensure AI labs remain in Australia to serve the global market.
• Pragmatic Regulation: We favour Input Taxes over Usage Taxes for local hardware to avoid the "Panopticon" of state-monitored processing.
• Human Centricity: We treat human employment as a necessary form of social integration. The tax system must be used to ensure that a booming AI economy does not result in a hollowed-out society.
• Sovereign Stockpiling: The "announcement effect" of this tax is a feature, not a bug. It will trigger a massive influx of compute capacity into Australia before the policy commences, providing the productivity backbone for our future GDP.