Growth, molting, and development of larval Antarctic krill were investigated near Marguerite Bay during four cruises in austral autumn and winter 2001 and 2002, as part of the US Southern Ocean GLOBEC program. Overwintering survival of larvae has been linked to annual sea-ice formation and extent, as sea-ice biota may provide food when other sources are scarce in the water column. During autumn 2001, larvae were very abundant (1–19 individuals m⁻³), with younger stages dominant offshelf and older stages dominant on-shelf. On-shelf larvae were in better condition than offshore larvae. During autumn 2002, larvae again were abundant offshelf (0.01–110 m⁻³), whereas all stages were scarce on-shelf. Declining diatom and radiolarian blooms were present during autumn in both years. Average chlorophyll concentrations were low (0.10 vs. 0.22 μg l⁻¹) in autumn and an order of magnitude lower in winter. Carbon content of larvae during autumn 2001 and 2002 (41% vs. 38% C of DW) suggested that lipid storage was moderate. The median autumn larval growth rate (0.027 mm d⁻¹) was lower and the intermolt period (19 d) longer than reported summer values. During winter, larvae appeared to be food-limited based on the following observations: (1) the median growth rate decreased (0.00 mm d⁻¹) and the intermolt period increased (40 d), (2) larval length-specific dry weight (DW) and % carbon and nitrogen of DW decreased, and (3) 88% of furcilia 6 did not develop to the juvenile stage, but remained at the same stage after molting. Experimental results demonstrated that some larvae could survive starvation for a month by combusting body reserves (ca. 1% decrease in DW and body C and N d⁻¹), implying that a portion of the population was resilient to the suboptimal food supply. Although sea ice formed up to 2 months earlier in 2002, ice algae at the ice–water interface, where it is accessible to krill, was not an abundant food source in either year (0.05 vs. 0.07 μg chl l⁻¹). In winter 2001, furcilia were commonly observed near the undersurface of sea-ice, but only rarely in 2002 until mid-September, when ice algae began to accumulate. Low gut fluorescence values also indicate that little nutrition was derived from autotrophs in winter. Instead, larvae were likely opportunistic scavengers exploiting all available food sources, including microzooplankton, benthic larvae, detritus, scarce phytoplankton and sea-ice biota. In summary, larval krill exhibited several overwintering behaviors: (1) flexible feeding, (2) flexible morphology (i.e., delayed development), (3) flexible physiology (i.e., increased intermolt period, reduced growth), (4) moderate lipid storage, and (5) ability to withstand starvation by combusting body C and N. Because most larvae did not shrink in length, this measure may not be a good indicator of the body combustion strategy. At these high latitudes, sea-ice biota may not be a primary source of food during winter, but progressively more important in spring as irradiance levels increase. Winter survivors during 2001 resulted in a significant recruitment to the juvenile size class during spring.